INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP BETWEEN FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH: EVIDENCE FROM TAIWAN AND JAPAN

Transcription

1 JOURNAL OF ECONOMIC DEVELOPMENT 49 Volume 30, Number 1, June 005 INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP BETWEEN FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH: EVIDENCE FROM TAIWAN AND JAPAN CHIEN-CHIANG LEE AND SWEE YOONG WONG * Naional Chung Hsing Universiy and Shih Chien Universiy This paper employs a hreshold regression model o invesigae he exisence of inflaion hreshold effecs in he relaionship beween financial developmen and economic growh. A specific quesion ha is addressed in his paper is wha he hreshold inflaion raes are for Taiwan and Japan. Resuls indicae ha here is one inflaion hreshold value in Taiwan, whereas here are wo in Japan. Earlier sudies suppor he view ha financial developmen may promoe economic growh. However, he conclusion drawn from he empirical findings suggess ha i can only be achieved under low and moderae inflaion. In addiion, he hreshold level of inflaion below which financial developmen significanly promoes growh is esimaed a 7.5% for Taiwan and 9.66% for Japan. The empirical findings from he hreshold regression model indicae ha inflaionary hreshold for boh counries occurred in he high inflaion period of he world energy crises in he 70s. Keywords: Threshold Regression Model, Inflaion, Financial Developmen, Economic Growh JEL classificaion: E44, O16, C 1. INTRODUCTION This paper invesigaes he possible exisence of inflaionary hreshold effecs in he relaionship beween financial secor developmen and economic growh, using a hreshold regression approach. Convenional hinking in macroeconomics holds ha, in general, low inflaion in conjuncion wih financial developmen is an essenial elemen in promoing economic growh. No surprisingly, he quesion of he exisence and naure of he link beween inflaion, financial developmen, and economic growh has * The auhors would like o hank he referees for heir valuable suggesions and helpful commens which have grealy enhanced he qualiy of his paper. Any remaining errors are, of course, belong o he auhors.

2 50 CHIEN-CHIANG LEE AND SWEE YOONG WONG been he subjec of considerable ineres and debae. Alhough he debae abou he precise relaionship beween hese hree variables remains open, an exhausive lieraure review on his issue has produced some imporan empirical findings and a relaively wide consensus abou some aspecs of he relaionship beween hese variables has been reached. In paricular, i is generally acceped ha inflaion has a negaive effec on medium- and long-erm growh due o he influence of he former on reducing invesmen and produciviy growh (Barro (1991); Fischer (1993); and Bullard and Keaing (1995)); inflaion impedes financial deepening (Bruno and Easerly (1998); Boyd e al. (001)); financial developmen plays an imporan role in he growh process (Diamond and Dybvig (1983); King and Levine (1993); and Levine (1997)); 1 inflaion reards economic growh hrough is negaive impac on he financial developmen (Azariadis and Smih (1996); Hyubens and Smih (1999)). If inflaion is a major obsacle in promoing economic growh,hen i readily follows ha policymakers should aim a a low rae of inflaion. However, wha should he arge level of inflaion be? How low should inflaion be in order for financial developmen o have a posiive impac on economic growh? In oher words, if a non-linear relaionship exiss beween inflaion and growh, hen i should be possible o esimae he hreshold, or inflexion poin, a which he sign of he relaionship beween he wo variables would swiches. The possibiliy of such a non-linear relaionship was firs idenified by Fischer (1993), who noed he exisence of a posiive relaionship beween long-run growh and inflaion a a low rae of inflaion, bu a higher raes i becomes negaive. Following he resuls of Fischer (1993), here is an expanding body of evidence indicaing ha he relaionship beween inflaion and long-run growh is characerized by non-lineariies and he exisence of hresholds (Sarel (1996); Bruno and Easerly (1998); and Singh and Kalirajan (003)). Going beyond he empirical invesigaion of he non-linear relaionship beween inflaion and growh, here are several heoreical and empirical sudies focusing specifically on he exisence of inflaionary hreshold effecs in he relaionship beween financial developmen and growh. Empirical and heoreical lieraure nowadays suggess ha financial markes play an imporan role in he growh process. Thus, if changes in he rae of inflaion do affec aciviy in financial markes, hen i is likely ha 1 As sandard heory shows, a financial sysem preserves he efficiency of he paymen sysems, and hey conribue o saving allocaion. Tha is, a more developed financial secor leads o a beer allocaion of resources, beer monioring, and less informaion asymmeries. A comprehensive survey is provided by Levine (1997). In Sarel (1996) sudy, when inflaion is below 8%, here is a posiive correlaion and vice versa. Bruno and Easerly (1998) choose he value 40% o be he breakpoin beween low and high inflaion regimes. Singh and Kalirajan (003) addressed he hreshold inflaion rae for India, bu heir findings show ha here is no significan srucural break in he inflaion-growh nexus in India. Inflaion has a significan negaive effec a all levels.

3 INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP 51 such changes would also have implicaions for long-run real aciviies. The sudies of Hyubens and Smih (1999), Gylfason and Herbersson (001), Bose (00), and Rousseau and Wachel (00) have demonsraed ha he level of inflaion is an imporan facor in affecing he relaionship beween financial developmen and growh. Moreover, hese sudies provide empirical evidence indicaing ha under a low or moderae inflaion rae, financial developmen promoes economic growh. On he conrary, under higher inflaion environmens, financial developmen does no have any impac on economic growh. The reason for his is ha he flow of informaion abou invesmen projecs and reurns being used by inermediaries becomes more uncerain and less readily available in an inflaionary environmen. Inflaion increases ransacions and informaion coss which direcly inhibi economic developmen. High inflaion can repress financial inermediaion by eroding he usefulness of money asses and by leading o policy decisions ha disor he financial srucure. Taiwan and Japan have gone hrough differen sages of developmen in he process of growh, however, here are some similariies beween he wo. Inernaional rade beween boh counries is closely ied, and hey have experienced rapid economic growh afer World War II. Banking secor plays an imporan role in he financial sysem and he process of growh in boh counries. In order o coninue inegraing ino he world economy and o enhance he abiliy of he financial sysem o mach he increasing demands of indusrializaion and exernal rade, boh counries lifed financial conrol and undergone a series of financial deregulaion in he 80s. Boh counries have experienced asse price boomed in he lae 80s, followed by a declining in asse value in he early 90s. According o he ime-series daa, here was several high inflaion periods occurred in he 70s and 80s in boh counries. The high inflaion periods in boh counries can be aribued o he world energy crises in he 70s and he asse price boomed in he 80s. Alhough exensive empirical work has been concenraed on he naure of he relaionship among inflaion, financial developmen, and economic growh, surprisingly, he lieraure on his subjec has largely negleced hese wo fas growing Asian economies. Even hough only a handful of sudies focus on his subjec for Taiwan and Japan (Fang (1995); Sinha and Macri (001); and Hou and Hu (00)), he possibiliy of he exisence of a non-linear relaionship beween hese hree variables has largely been ignored in hese sudies. In ligh of he heoreical and ad-hoc empirical lieraure, his paper re-examines he above issue for Taiwan and Japan wih an emphasis specifically on he following quesions: 1. Wha are he effecs of financial developmen on economic growh under differen inflaion raes?. Is here a saisical significan hreshold level of inflaion above which inflaion is derimenal o financial developmen? 3. Is here a saisically significan non-linear relaionship beween inflaion and growh; in oher words, is here a hreshold level of inflaion above which inflaion affecs growh differenly han a lower inflaion raes?

4 5 CHIEN-CHIANG LEE AND SWEE YOONG WONG 4. Considering he differen hreshold esimaes obained in he sudies of Gylfason and Herbersson (001) and Rousseau and Wachel (00), are hese hreshold values differen from hose obained in his sudy? 3 These quesions are examined using new economeric mehods for hreshold esimaion and inference. The resuls obained srongly suppor inflaionary hreshold effecs in he relaionship beween financial developmen and growh in Taiwan and Japan. The organizaion of he paper is as follows. Secion discusses research mehodology. Secion 3 provides daa descripion and empirical resuls followed by a concise background of Taiwan and Japan economies. Concluding remarks and policy implicaions are made in Secion 4.. METHODOLOGY AND IMPLEMENTATION There is a paricular economeric issue relaed o he esimaion and inference in models wih hreshold effecs. Mehods need o be developed o conduc empirical esimaion and inference. Toward his regard, he empirical mehodology used in his sudy is based on he hreshold auoregressive (hereafer TAR) approach inroduced by Tong (1978) and Hansen (1996). This model specifies ha individual observaions can fall ino discree classes based on he value of an observed (hreshold) variable. In growh heory, he main sources of growh power lie in he accumulaion of he facor of producion, and he promoion of marginal produciviy and oal facor produciviy. 4 Based on he framework of Odedokun (1996), he producion funcion incorporaes financial developmen as one of he inpu facors given as: Y = F L, K, FD ; Z ), (1) ( where Y denoes he real oupu or real GDP; L denoes he labor force; denoes he capial sock; FD denoes he financial developmen level; Z denoes he oher facors ha affec he aggregae hroughpu. Afer linearizing Equaion (1) we obain he following empirical equaion: Y & = β & & & Z& + u, () 0 + β1l + βk + β3fd + β4 K 3 Gylfason and Herbersson (001) applied he random-effecs panel model; he resuls show ha inflaion in excess of 10-0 percen per year is generally derimenal o growh. Rousseau and Wachel (00) employed rolling panel regressions showing ha here is an inflaion hreshold for he finance-growh relaionship ha lies beween 13 and 5 percen. 4 For a deail explanaion, please see Chaper Ten of Barro and Sala-i-Marin (004).

5 INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP 53 where Y &, L &, K &, F D &, and Z & respecively denoe real GDP growh rae, labor force growh rae, invesmen rae, financial developmen index, and oher facors, and u denoes he error erm. Following he empirical framework of Beck e al. (000) and Chrisopoulos and Tsionas (004), we include expor growh rae ( E X & ) and inflaion rae ( P & ) o serve as he oher facors ha affec economic growh. Thus, he empirical regression equaion can be se as: Y & = δ + δ L& + δ K& + δ FD& + δ EX& + δ P& + u. (3) * According o he growh heory, and δ are boh posiive coefficiens given δ1 ha labor force and capial sock have a posiive impac on he aggregae oupu. The sign and he saisical significance of he regression coefficien measure he effecs and he significan impac of financial developmen on economic growh. From he growh-heory lieraure poin of view, represens posiive coefficiens given ha expor expansion promoes economic growh. The regression coefficien δ 4 reflecs he impac of inflaion on economic growh. The sudies of De Gregorio (199) and Jones and Manuelli (1993) provide evidence suggesing ha inflaion has a significanly negaive impac on economic growh. However, he sudy of Khan and Senhadji (001) suggess ha here is a hreshold level of inflaion above which inflaion is derimenal o economic growh, and below his paricular level, inflaion is posiively relaed o growh. Given he conradicory resuls found in he foregoing sudies, he sign of he regression coefficien is uncerain; herefore, i is subjec o an invesigaion in his δ 5 sudy. The empirical Equaion (3) represens he convenional linear growh model. However, recen lieraure has found ha he relaionship beween financial developmen and economic growh does no follow a single paern. For example, Huybens and Smih (1999), Gylfason and Herbersson (001), and Bose (00) have found ha inflaion affecs real variables hrough is impac on he financial marke aciviies. In order o examine he effecs of he ineracion beween inflaion and financial developmen on economic growh, we employ he TAR model proposed by Tong (1978) and Hansen (1996). In he TAR model, he classificaion of he variable across regimes is based on an esimae of he ime series behavior ha is consisen wih reaching he hreshold ha separaes he regimes. Consider a wo-regime TAR model proposed by Hansen (1996): ' y = θ x + e if q γ, (4) y 1 1 = ' θ x + e if q > γ, (5) δ 3 δ 5

6 54 CHIEN-CHIANG LEE AND SWEE YOONG WONG where q denoes he hreshold variable, pariioning all he observed values ino wo groups or regimes. Term y denoes he variable o be explained, whereas x is a marix ha, denoes he explanaory variable. The error erm e i is whie-noise iid, and γ represens he hreshold value, which is unknown ex ane, bu can be esimaed. The foregoing model shows ha when he hreshold variable is smaller han he hreshold parameer, he regression Equaion (4) is applicable. When he hreshold variable is larger han he hreshold parameer, he regression Equaion (5) is applicable. Le I ( γ ) = { q γ}, and { } as an indicaor funcion wih I = 1 if q γ occurs, or I = 0 oherwise. Moreover, when we le x ( γ ) = x I ( γ ), hen Equaions (4) and (5) can be revised as follows: y = θ ' x + ρ' x γ ) + e, e ~ iid(0, σ ). (6) ( Therein, θ = θ, ρ = θ 1 θ, e = [ e1 e ]', and θ, ρ, and γ are he parameers o be esimaed. Equaion (6) allows all he regression coefficiens o differ beween sample groups. The resuling sum of squared error as a resul of esimaing hese parameers θ, ρ, and γ can be expressed as follows: S ( γ ) = eˆ ( γ )' eˆ ( ). (7) 1 γ Hansen (1996) recommended an esimaion of γ by he leas-squares mehod. This can be achieved by minimizing he sum of squared errors in (7). Hence, he opimum hreshold value is given as: ˆ γ = arg min S ( ), (8) 1 γ and he variance of he residual is expressed as: 1 1 ˆ σ = eˆ ˆ e = S1( ˆ) γ. (9) T T Once γˆ is obained, he vecors of he slope coefficien esimaed are ˆ θ = ˆ( θ ˆ) γ and ˆρ = ˆ ρ( ˆ γ ). According o he foregoing process, he regression Equaion (3) under a wo-regime TAR model can be expressed as: Y& = ( δ + ( δ δ11l& + δ1k& + δ13fd& + δ14ex& + δ15p& ) I[ q γ ] * + δ L& + δ K& + δ FD& + δ EX& + δ P& ) I[ q > γ ] + e (10)

7 INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP 55 In he esimaion of model (10), he hreshold value is deermined by obaining he hreshold value ha minimizes he sum of he squared error given by Equaion (7). Since he purpose of his paper is o invesigae he inflaionary hreshold effecs in he relaionship beween financial developmen and economic growh, he annual growh rae of inflaion is employed as he main hreshold variable in he analysis. Before applying he wo-regime TAR model such as (10), we need o es for he hreshold effecs. In his case, i is analogous o esing he null hypohesis of he linear model agains he alernaive hypohesis of he wo-regime model such as equaion (10). Convenional mehods of hypohesis esing canno be applied, because of he difficuly wih he hreshold parameer γ being unidenified under he null hypohesis. In his case, he underlying disribuion of a large sample is no disribued by Chi-square disribuion. Hansen (000) proposed a soluion for his paricular problem and suggesed ha relevan ess be conduced hrough he use of a Lagrange Muliplier (LM) boosrap procedure. Since γ is no idenified under he null hypohesis of he no-hreshold effec, he p-values are compued by a fixed boosrap mehod. To examine by esing wheher he coefficiens in he wo regimes are he same or no, he null hypohesis of no hreshold effec o Equaion (10) is: H0 : 1i δi δ = i = (11) Le and S be he residual sum of squares under he null hypohesis and S0 1 alernaive of (11). As such, he F -es is based on: F S S ( ˆ) γ ˆ σ =. (1) Once he hreshold effec exiss, he nex quesion is wheher or no he hreshold value can be known. The null hypohesis of he hreshold value is H0 : γ = γ 0, and he likelihood raio saisics is: LR ( γ ) 1 S ( γ ) S ( ˆ) γ ˆ σ 1 1 =, (13) where S ( ) and S 1 ( ˆ γ ) are he residual sum of squares from Equaion (7) given he 1 γ rue and esimaed value, respecively. The asympoic disribuion of LR 1 ( γ 0 ) can be used o form valid asympoic confidence inerval abou he esimaed hreshold values. The saisics of LR 1 ( γ 0 ) are no normally disribued and Hansen(000)compued

8 56 CHIEN-CHIANG LEE AND SWEE YOONG WONG heir no-rejecion region, c (α ), α is a given asympoic level. 5 Tha is, if LR 1 ( γ 0 ) c( α), where c( α) = ln( 1 1 α ), hen he null hypohesis of H 0 : γ = γ 0 6 canno be rejeced. Aside from esing he exisence of one hreshold value, o furher invesigae wheher here are wo or more hreshold values ha exis, we firs employ he es o assess he null hypohesis of no hreshold. If his null hypohesis is rejeced, hen a leas one hreshold value is ensured. We nex proceed o es he null of one hreshold agains he wo hresholds. We assume a known esimaed and proceed o search he second hreshold, γ. In his case, we obain he following: γˆ1 F 1 ( γ ) = S( ˆ γ 1, γ ) S( γ, ˆ γ ) S r 1 if ˆ γ < γ 1 if γ < ˆ1. γ (14) The hreshold value, he null hypohesis, and he follows: F -es are respecively saed as γ r r = arg min S ( ), (15) ˆ : H 0 γ only one hreshold, (16) S1(γˆ 1) S F = ˆ σ ( ˆ ), (17) r γ r where S 1 ˆ γ ) ( 1 1 σ = S T r ˆ is referred o as he sum of squared errors acquired from he previous hreshold esimaion. The residual variance is given as follows: r ( γ ). (18) ˆ The significan F implies he rejecion of he null of one hreshold and wo hresholds are expeced. If wo hresholds canno be rejeced, han he confidence inerval for wo hresholds ( γ 1,γ ) can be consruced in he same way. The procedures are carried ou unil he null in (11) can no longer be rejeced. 5 The no-rejecion region of he confidence level is ( 1 α). 6 Equaions (13) and (1) are differen, and LR ( ) ess if he hreshold esimaor γˆ is equal o acual hreshold regime γ 0 1 γ, and he F disribuion ess if he coefficiens of he wo regimes are equal or no.

9 INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP THE EMPIRICAL RESULTS 3.1. The Descripion of he Daa Se and Variables The principle daa source of Taiwan in his paper is aken from he AREMOS economic saisics daabase, which is a public daabase compiled by he Minisry of Educaion, Taiwan, R.O.C. Japan s daa all emanae from IMF s Inernaional Financial Saisics. The daase for Taiwan sars from he firs quarer of 1965 o he las quarer of 00 wih 001 serving as he base year for he Consumer Price Index (CPI). The daase for Japan sars from he firs quarer of 1970 o he las quarer of 001 wih 1995 as he base year for CPI. The descripion of he variables is as follows: Y & = The economic growh rae represened by he annual growh rae of gross domesic produc (GDP). L & = The populaion growh rae has been used as he proxy for labor force growh which represened by he annual growh rae of oal populaion. K & = Real invesmen variable represened by he annual growh rae of real gross domesic fixed capial formaion. F D & = Financial developmen variable represened by he annual growh rae of real liquid liabiliies of he financial secor (currency plus demand and ineres-bearing liabiliies of banks and non-bank financial inermediaries) as he index for financial deph according o Odedokun (1996). 7 E X & = Real expor variable represened by he annual growh rae of real commodiies and services expor. The inclusion of his variable is o measure he degree of rade openness. Moreover, given he fac ha Taiwan s economy is expor-oriened, is performance in rade has a profound impac on he domesic economy. P & = Inflaion variable represened by he annual growh rae of CPI. Table 1 repors some descripive saisics of he six key variables. Taiwan and Japan have he highes inflaion of 61.4% and 3.4%, respecively, whereas Taiwan s average inflaion is 5.1%, and Japan is only around 3.8%. Table presens he correlaion marix of explanaory variables for Taiwan. As can be seen, he highes correlaion is 0.65 beween he inflaion rae and finance developmen indicaor and he lowes correlaion is 0.01 beween he inflaion rae and invesmen rae. The remaining correlaion coefficiens are around o 0.40, which are accepable o avoid he muli-collineariy problem. Table 3 presens he correlaion marix of explanaory variables for Japan, where he highes correlaion is beween he invesmen rae 7 This indicaor measures he overall size of he formal financial inermediary secor, and has been found o be very srongly associaed wih boh he level and economic growh rae.

10 58 CHIEN-CHIANG LEE AND SWEE YOONG WONG and finance developmen indicaor and he lowes correlaion is beween he inflaion rae and populaion growh. The remaining correlaion coefficiens are around -0. o 0.455, which are accepable o avoid he muli-collineariy problem. Table 1. Summary Saisics of Variables Variables Mean Sd. Dev. Max. Min. Skewness Kurosis Taiwan Y & L & K & F D E X P & Japan Y & L & K & F D E X P & Table. Correlaion Marix: Taiwan L & K & F D E X P & L & 1 K & F D E X P & Table 3. Correlaion Marix: Japan L & K & F D E X P & L & 1 K & F D E X P &

11 INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP Tes for he Exisence of Inflaionary Threshold Effecs The Phillips and Perron (PP; see Phillips and Perron (1988)) es is employed o examine he hypohesis of a uni roo among he variables. The resuls of Taiwan and Japan are presened in Table 4. In each case, he number of lags is chosen such ha no serial auocorrelaion in residuals is eviden in he auxiliary regressions. The resuls of he es show ha he null hypohesis of uni roo is rejeced a he 5% level of significance in drif erm or drif erm and ime rend, which implies ha he underlying variables for Taiwan and Japan show a saionary process. Table 4. Resuls of Uni Roo Tess Counry Taiwan Variable Drif erm Drif erm and ime rend Y & ** ** L & ** K & -4.61** -5.55** F D ** -4.75** E X ** -6.16** P & ** ** Counry Japan Variable Drif erm Drif erm and ime rend Y & ** ** L & -.883** K & -3.5** ** F D -3.65** ** E X -4.14** ** P & ** Noes: The 5% criical values wih ime rend esimaion and wihou ime rend esimaion are respecively and The criical values are from Mackinnon (1991). ** indicaes ha a he 5% level of significance. Before applying he TAR model such as Equaion (10) o invesigae he impac of he financial developmen on economic growh under differen inflaion regimes in Taiwan and Japan, we proceed o es for he hreshold effecs wih he annual growh rae of inflaion serving as he hreshold variable. In his case, i simply amouns o esing he null hypohesis of he linear model agains he alernaive hypohesis of he wo-regime model. Because of he hreshold parameer γ being unidenified under he null hypohesis, convenional ess (such as he -es) have non-sandard disribuions. Hansen (1996) proposed a boosrap mehod o simulae he asympoic disribuion of

12 60 CHIEN-CHIANG LEE AND SWEE YOONG WONG he likelihood raio es of he null hypohesis. The resuls of he hreshold es and he asympoic p-values obained hrough 1,000 boosrap replicaions are repored in Table 5 and Table 6 for Taiwan and Japan, respecively. Table 5. Tes Resuls of Inflaion Threshold Effec and Threshold Esimaes for Taiwan Null of hypohesis F es Boosrap p-value Esimae hreshold Null of no hreshold 16.63** 0.09 ˆ1 γ = 7.56% Null of one hreshold Noes: Esimaion period is 1965Q1 o 00Q4. Threshold variable is denoed as p&. The hresholds are obained by he minimum sum of he squared residual. The F es is calculaed based on (13). ** is significan for 5%. Table 6. Tes Resuls of Inflaion and Threshold Effec Threshold Esimaes for Japan Null of hypohesis F es Boosrap p-value Esimae hreshold Null of no hreshold 6.046** Null of one hreshold ** Null of wo hresholds γ =.578% r ˆ1 r ˆ γ = % Noes: Esimaion period is 1970Q1 o 001Q4. Threshold variable is denoed as P &. The hresholds are obained by he minimum sum of he squared residual. The F es is calculaed based on (13). ** is significan for 5%. The hreshold effec is firs examined. We employ he F- es o examine he equaliy of coefficiens in he wo regimes, and he LR es examines he equaliy of poenial hreshold values. Table 5 repors he esing resuls and he boosrapped p-values for Taiwan. The saisics of F1 is for specificaion (10), as F1 exceeds he criical values a he 5% significan level, suggesing one hreshold a leas. Then we proceed o invesigae he possibiliy of he exisence of more han one hreshold, and he es resuls canno rejec he one hreshold s null hypohesis. This implies ha one inflaion hreshold value exiss for Taiwan, which is 7.56%. As a resul, he daa on he growh rae of inflaion for Taiwan can be divided ino low and high inflaion regimes. The esimaion resuls are quie similar o he resuls repored in he sudies of Sarel (1996) and Khan and Senhadji (001). Sarel (1996) invesigaed he non-linear relaionship beween inflaion and economic growh. A panel daa analysis was employed o analyze a sample of 87 counries covering he period from 1970 o The resul showed ha here was a remarkable srucural hreshold value beween inflaion and economic growh; he hreshold value was 8%. Khan and Senhadji (001) provided evidence o sugges ha he inflaion hreshold value for indusrial counries was in beween 1% o 3%; in he case of developing counries, i was in-beween 7% o

13 INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP 61 11%. Once he hresholds are obained, he nex sep is compuing he LR es o examine heir respecive confidence inerval. The 95% asympoic confidence inerval as [7.05%, 7.60%]. This exremely igh confidence inerval suggess ha he hreshold esimaes are very precise. Hence, here is srong evidence supporing one hreshold in he regression relaionship. Compared wih his sudy, boh of he previous sudies did no invesigae he impac of he relaionship beween inflaion and financial developmen on economic growh. In addiion, hese sudies employed pooled cross secional ime-series daa, whereas his sudy considers each counry individually by employing ime-series daa. Table 6 repors ha he saisics F1 and F exceed he boosrapped criical values a he 5% significan level for Japan, suggesing he possibiliy of wo hresholds, which are.578% and %. Consequenly, he daa on he growh rae of inflaion for Japan can be divided ino low, moderae, and high inflaion regimes. We compue he LR es menioned above, and he 95% confidence inerval as [.44%,.73%] and [8.1%, 9.96%] for he above wo hresholds, respecively Inflaion, Financial Developmen, and Economic Growh Once he search for he inflaion hreshold effecs is conduced for Taiwan and Japan, we proceed o invesigae he relaionship beween inflaion, financial developmen, and economic growh under differen inflaionary regimes for boh counries. Table 7. The Esimaion Resuls of he Inflaion Financial Developmen and Economic Growh for Taiwan Explanaory Variables Linear model Low inflaion rae High inflaion rae Threshold Value 7.56% > 7.56% Consan (.800)** L & (-0.379) K & 0.10 (5.175)** F D & 0.37 (5.861)** E X & (7.34)** P & (-0.91) (1.874)* (-0.663) (4.816)** 0.97 (7.115)** (5.90)** (-0.186) 0.05 (0.673) (0.04) (1.708)* (0.64) (8.385)** (-3.413)** R Numbers of Sample Noes: Esimaion period is 1965Q1 o 00Q4. Threshold variable is P &. ** and * respecively represen 5% and 10% levels of significance. Values in parenheses are values.

14 6 CHIEN-CHIANG LEE AND SWEE YOONG WONG Table 7 summarizes he esimaion resuls of he relaionship beween hese hree variables for Taiwan. For comparison purposes, he firs column displays esimaes for a linear specificaion ha ignores he hreshold effec. The remaining columns provide esimaes of he TAR model. The resuls of he firs column indicae ha financial developmen is an imporan ingredien in promoing Taiwan s economic growh, as he significan coefficien is Table 7 shows ha when he annual growh rae of inflaion is below 7.56%, financial developmen has a significan effec on economic growh in Taiwan, where he coefficien is However, when inflaion is above his hreshold value, financial developmen does no have any significan impac on economic growh. As a resul, financial developmen ha promoes economic growh can only be esablished under low inflaion. This empirical finding is consisen wih he conclusion derived from he heoreical models of Huybens and Smih (1999), and Bose (00). Noe ha boh invesmen and expor variables have a posiive and significan impac on economic growh regardless of he inflaion regimes. This finding reinforces he fac ha Taiwan s economy indeed heavily depends on rade. Table 8. Explanaory Variables Threshold Values The Esimaion Resuls of he Inflaion Financial Developmen and Economic Growh for Japan Linear model Low inflaion rae Moderae inflaion High inflaion rae rae.578%.578% ~ % % Consan (4.198)** (6.838)** (-3.53)** 0.08 (.390)** L & (3.83)** (7.634)** 0. (1.844)* (0.93) K & 0.78 (11.393)** 0.54 (10.477)** 0.36 (7.705)** 0.6 (1.86)* F D (6.11)** (3.605)** (5.530)** (-0.839) E X (-0.875) (1.13) (-4.10)** (5.48)** P & (0.776) (-1.45) 0.58 (.997)** (-.836)** R Numbers of Sample Noes: Esimaion period is 1970Q1 o 001Q4. Threshold Variable is P &. ** and * respecively represen 5% and 10% levels of significance. Values in parenheses are values.

15 INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP 63 Table 8 summarizes he esimaion resuls of he relaionship beween inflaion, financial developmen, and economic growh for Japan. The empirical resuls obained from he esimaion of he linear model show ha financial developmen is an imporan ingredien in promoing Japan s economic growh, as he significan coefficien is 0.3. The non-linear esimaion produces resuls suggesing ha under a low (lower han.578%) or moderae (beween.578% and %) inflaion rae, financial developmen has a significanly posiive impac on economic growh. However, under a high inflaion regime (inflaion rae higher han %), financial developmen is unrelaed o economic growh. Due o he resul, he conclusion ha financial developmen may promoe economic growh can be esablished only when Japan s inflaion rae is low or moderae The Relaionship Beween Inflaion and Economic Growh Table 7 also provides empirical evidence ha in Taiwan, inflaion has no significan negaive impac on economic growh in linear esimaions. However, under a high inflaion regime (inflaion above he hreshold value of 7.56%), inflaion is derimenal o economic growh, as he significan coefficien is In he case of Japan, he resuls summarized in Table 8 indicae ha under a moderae inflaion regime (inflaion rae in beween.578% and %), inflaion promoes economic growh, where he significan coefficien is However, when inflaion is higher han %, inflaion has a significan negaive effec on economic growh, as he coefficien is The foregoing esimaed non-linear relaionship beween inflaion and economic growh is consisen wih he empirical and heoreical conclusion derived in he sudies of Sarel (1996), and Bose (00); ha is, high inflaion has a negaive effec on economic growh. This paper addresses wha he hreshold inflaion raes are for Taiwan and Japan Inflaionary Thresholds and he Economies of Taiwan and Japan Afer World War II, Taiwan and Japan achieved remarkable record of high and susained growh, ofen in he conex of acivis public policies. Beween 1960 and1985, real income per capia increased more han four imes in Japan and Taiwan, and more han double in he Souheas Asian Newly Indusrialized Naions (NIEs). 8 The specacular economic performance in Taiwan is aribuable o a series of growh-enhancing public policy wih low inflaion and macroeconomic sabiliy as a framework, and he developmen and liberalizaion of he financial sysem. Figure 1 summarizes he economic performance, inflaion, and he financial developmen in 8 See Figure, page 3 of The Eas Asian Miracle: Economic Growh and Public Policy, A World Bank Policy Research Repor, 1993.

16 64 CHIEN-CHIANG LEE AND SWEE YOONG WONG Taiwan over he sample period. I shows ha economic growh in Taiwan is highly correlaed wih is financial developmen. The developmen policy of Taiwan has consised of five sages in which he governmen has implemened comprehensively. Namely, hey are Land Reform and Reconsrucion (1949-5), Impor-Subsiuing Indusrializaion ( ), Expor Promoion (1958-7), Indusrial Consolidaion and New Expor Growh ( ), and High Technology and Modernizaion (1981-). 9 The financial sysem in Taiwan has ofen been criicized as being underdeveloped and inefficien, paricularly, he financial sysem s inabiliy o mach he increasing demands of indusrializaion and exernal rade. The financial sysem in Taiwan was even regarded as a major obsacle o furher economic growh in he 80s. As a resul, he governmen launches a series of financial liberalizaion (which include deregulaion and inernaionalizaion) during 1980s. 10 In he lae 1980s and early 1990s, sock and real esae markes boomed, and he economy experienced he so-call bubble economy. Figures 1 and provide furher evidence o sugges ha in he process of economic growh, high inflaion period of he 70s resuling from he world energy crises has a profound impac on he financial developmen and economic growh for boh counries. The empirical findings from he TAR analysis indicae ha inflaionary hresholds for boh counries occurred in he high inflaion period of he 70s. In he case of Taiwan, inflaionary hreshold value of 7.5% occurring in he hird quarer of As for Japan, i s inflaionary hreshold value is 9.66%, occurring in he hird quarer of The shaded areas in Figures 1 and represen he period of high inflaion for Taiwan and Japan respecively. Boh figures show ha high inflaion in he period from 1973 o 1974 had he sronges negaive effec on he financial developmen and economic growh for boh counries. For example, in he case of Taiwan, when inflaion reaches is highes level of 61.41% in he firs quarer of 1974, he degree of financial developmen and economic growh were he lowes, hey were, -3.13% in he firs quarer of 1974 and -14.0% in he las quarer of 1974, respecively. As for Japan, he highes inflaion rae during he period from 1973 o 1974 was 3.45% in he hird quarer of 1974, and lowes degree of financial developmen and economic growh were -7.69% in he hird quarer of 1974 and -4.66% in he las quarer of 1974, respecively. Boh counries experience low and moderae inflaion afer 1981, and inflaion was generally sable in he 90s. 9 The Eas Asian Miracle: Economic Growh and Public Policy, A World Bank Policy Research Repor, 1993, pp The financial liberalizaion process in Taiwan could be divided in hree sages. Firs, ineres rae liberalizaion sared in he early 1980s; second, he foreign exchange sysem was convered from a fixed rae sysem o a managemen flexible rae sysem in Third, he liberalizaion of he securiies marke sared in 1988.

17 INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP % inflaion rae % financial developmen % economic growh rae Figure 1. Plo of Inflaion Rae, Financial Developmen and Economic Growh by Inflaion Threshold for Taiwan

18 66 CHIEN-CHIANG LEE AND SWEE YOONG WONG 5 0 % inflaion rae % financial developmen % economic grow rae Figure. Plo of Inflaion Rae, Financial Developmen and Economic Growh by Inflaion Threshold for Japan

19 INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP CONCLUSIONS AND POLICY IMPLICATIONS According o he heoreical sudies of Huybens and Smih (1999), Gylfason and Herbersson (001), and Bose (00), inflaion will affec real economic aciviies hrough is impac on he financial marke. De Gregorior and Guidoi (1995) also produced similar resuls for Lain American counries. Tha is, when Lain America counries underwen high inflaion rae in he 1970s and 1980s, financial developmen rearded economic growh. The aim of his sudy is o broaden he empirical lieraure on he relaionship beween inflaion, financial developmen, and economic growh by invesigaing he relaionship among hese hree variables for Taiwan and Japan. Toward his regard, we employed he TAR approach inroduced by Tong (1978) and Hansen (1996) o invesigae he possible inflaionary hreshold effecs in he relaionship beween financial developmen and economic growh. The empirical findings of his sudy show ha here is one inflaion hreshold value in Taiwan, whereas here are wo in Japan. By reaing he annual growh rae of inflaion as he hreshold variable and exploring he effecs of financial developmen on economic growh under differen inflaion regimes, he esimaion resuls sugges ha when he hreshold level of inflaion is below 7.5%, financial developmen may promoe economic growh for Taiwan. However, when inflaion is higher han 7.5%, financial developmen will no generae any significan impac on economic growh. Consequenly, financial developmen ha promoes economic growh can only be esablished under low inflaion. As for Japan, The empirical resuls sugges ha when he hreshold level of inflaion is below 9.66%, financial developmen has a significanly profound impac on economic growh. However, financial developmen is derimenal o economic growh when inflaion is above he hreshold level. As a resul, he conclusion ha financial developmen may promoe economic growh can be esablished only when Japan s inflaion rae is low or moderae. This conclusion is consisen wih he findings of Huybens and Smih (1999), Bose (00), and Rousseau and Wachel (00). The esimaed resuls obain in his sudy sugges ha inflaionary hreshold indeed exiss in he relaionship beween financial developmen and economic growh in Taiwan and Japan. This implies ha changes in inflaion conribue one of he facors ha cause srucural change in he relaionship beween financial developmen and economic growh. Therefore, his srucural change due o he exisence of inflaionary hreshold should be aken ino accoun when consrucing esimaion and predicion models of economic growh for Taiwan and Japan. Furhermore, he policy implicaion derived from his sudy is ha before policymakers adop any policy o promoe or accelerae financial developmen, he role of inflaion should no be negleced in he relaionship beween financial developmen and economic growh; oherwise, such a policy may be derimenal o economic growh.

20 68 CHIEN-CHIANG LEE AND SWEE YOONG WONG REFERENCES Azariadis, C., and B. Smih (1996), Privae informaion, money, and growh: Indeerminacy, flucuaions, and he Mundell-Tobin effec, Journal of Economic Growh, 1, Barro, R. (1991), Economic growh in a cross secion of counries, Quarerly Journal of Economics, 104(), Beck, T., R. Levine, and N. Loayza (000), Finance and he source of growh, Journal of Financial Economics, 58, Bose, B. (00), Inflaion, he credi marke, and economic growh, Oxford Economic Papers, 54, Boyd, J., R. Levine, and B. Smih (001), The impac of inflaion on financial secor performance, Journal of Moneary Economics, 47, Bruno, M., and W Easerly. (1998), Inflaion cries and long-run growh, Journal of Moneary Economics, 41, 3-6. Bullard, J., and D. Keaing (1995), The long-run relaionship beween inflaion and oupu in poswar economies, Journal of Moneary Economics, 36, Chrisopoulos, D.K., and E.G. Tsionas (004), Financial developmen and economic growh: evidence from panel uni roo and coinegraion ess, Journal of Developmen Economics, 73, De Gregorio (199), The effecs of inflaion on economic growh: Lessons from Lain America, European Economic Review, 36, De Gregorio., and P. Guidoi (1995), Financial developmen and economic growh, World Developmen, 3(3), Diamond, D., and P. Dybvig (1983), Bank runs, deposi insurance, and liquidiy, Journal of Poliical Economy, 91(3), Fang, W.-S. (1995), The sudy of Taiwan inflaion, Quarerly of he Bank of Taiwan, 46(3), Fischer, S. (1993), The role of macroeconomic facors in growh, Journal of Moneary Economics, 3, Gylfason, T., and T. Herbersson (001), Does inflaion maer for growh? Japan and he World Economy, 13, Hansen, B. (1996), Inference when a nuisance parameer is no idenified under he null hypohesis, Economerica, 64, (000), Sample Spliing and Threshold Esimaion, Economerica, 68:3, Hou, T.-C., and C.-T. Hu (00), The esablishmen of esimaion model for inflaion in Taiwan, Cenral Bank of China Quarerly Bullein, 4(3), Huybens, E., and B. Smih (1999), Inflaion, financial markes, and long-run real aciviy, Journal of Moneary Economics, 43, Jones, L., and R. Manuelli (1993), Growh and he effecs of inflaion, NBER Working Paper Series No.453.

21 INFLATIONARY THRESHOLD EFFECTS IN THE RELATIONSHIP 69 Khan, M., and A. Senhadji (001), Threshold effecs in he relaionship beween inflaion and growh, IMF Saff Papers, 48, 1-1. King, R., and Levine, R. (1993), Finance and growh: Schumpeer migh be righ, Quarerly Journal of Economics, 108(3), Levine, R. (1997), Financial developmen and economic growh: Views and agenda, Journal of Economic Lieraure, 35(), Odedokun, M. (1996), Alernaive economeric approaches for analyzing he role of he financial secor in economic growh: Time-series evidence from LDCs, Journal of Developmen Economics, 50, Parick, H.T., and Y.C. Park (1994), The Financial Developmen of Japan, Korea, and Taiwan: Growh, Repression, and Liberalizaion, Oxford Universiy Press, New York. Phillips, P., and P. Perron (1988), Tesing for a uni roo, Economerica, 75, Rousseau, P., and P. Wachel (00), Inflaion hresholds and he finance-growh nexus, Journal of Inernaional Money and Finance, 1, Sarel, M. (1996), Nonlinear effecs of inflaion on economic growh, IMF Saff Papers, 43, Shen, C.-H., and D. Hakes (1995), Moneary policy as a decision-making hierarchy: The case of Taiwan, Journal of Macroeconomics, 17(), Singh, K., and K. Kalirajan (003), The inflaion-growh nexus in India: An empirical analysis, Journal of Policy Modeling, 5, Sinha, D., and J. Macri (001), Financial developmen and economic growh: The case of eigh Asian counries, Economia Inernazionale, 54(), Tong, H. (1978), On a hreshold model: Paern recogniion and signal processing, Amserdan: Sijhoff and Noordhoff, (1983), Threshold Models in Non-Linear Time Series Analysis, Heidelberg: Springer-Verlag. World Bank (1993), The Eas Asian Miracle: Economic Growh and Public Policy, A World Bank Policy Research Repor, Oxford Universiy Press, New York, Mailing Address: Chien-Chiang Lee, Deparmen of Applied Economics Naional Chung Hsing Universiy 50 Kuo-Kuang Rd., Taichung, Taiwan 40, R.O.C. Tel: ex. 308, Fax: ccl@nchu.edu.w. Manuscrip received Augus, 004; final revision received March, 005.