Policy Analysis Unit (PAU) Working Paper Series: WP 0604

Transcription

1 [Cover page] Policy Analysis Uni (PAU) Working Paper Series: WP 0604 Inflaion and Economic Growh in Bangladesh: Shamim Ahmed Md. Golam Moraza December 2005 Policy Analysis Uni (PAU) Research Deparmen, Bangladesh Bank Head Office, Dhaka, Bangladesh ( (

2 Policy Analysis Uni (PAU) Working Paper Series: WP 0604 Inflaion and Economic Growh in Bangladesh: Shamim Ahmed Research Economis, Policy Analysis Uni Research Deparmen Bangladesh Bank Md. Golam Moraza Senior Research Associae Cenre for Policy Dialogue December 2005 Copyrigh 2005 by Bangladesh Bank The Bangladesh Bank (BB), in cooperaion wih he World Bank Insiue (WBI), has formed he Policy Analysis Uni (PAU) wihin is Research Deparmen in July The aim behind his iniiaive is o upgrade he capaciy for research and policy analysis a BB. As par of is mandae PAU will publish, among oher, several Working Papers on macroeconomic research compleed by is saff every quarer. The precise opics of hese papers are chosen by he Residen Economic Adviser in consulaion wih he PAU members. These papers reflec research in progress, and as such commens are mos welcome. I is anicipaed ha a majoriy of hese papers will evenually be published in learned journals afer he due review process. Neiher he Board of Direcors nor he managemen of Bangladesh Bank, nor WBI, nor any agency of he Governmen of Bangladesh or he World Bank Group necessarily endorses any or all of he views expressed in hese papers. The laer reflec views based on professional analysis carried ou by he saff of he Policy Analysis Uni, and hence he usual cavea as o he veraciy of research repors applies. [An elecronic version of his paper is available a i

3 Inflaion and Economic Growh in Bangladesh: Shamim Ahmed and Md. Golam Moraza December 2005 Absrac I is widely believed ha moderae and sable inflaion raes promoe he developmen process of a counry, and hence economic growh. Moderae inflaion supplemens reurn o savers, enhances invesmen, and herefore, acceleraes economic growh of he counry. This paper empirically explores he presen relaionship beween inflaion and economic growh in he conex of Bangladesh. Using annual daa se on real GDP and CPI for he period of 1980 o 2005, an assessmen of empirical evidence has been acquired hrough he co-inegraion and error correcion models. Furher, i explores an ineresing policy issue of wha is he hreshold level of inflaion for he economy. The empirical evidence demonsraes ha here exiss a saisically significan long-run negaive relaionship beween inflaion and economic growh for he counry as indicaed by a saisically significan long-run negaive relaionship beween CPI and real GDP. In addiion, he esimaed hreshold model suggess 6- percen as he hreshold level (i.e., srucural break poin) of inflaion above which inflaion adversely affecs economic growh. These resuls have imporan policy implicaions for boh domesic policy makers and he developmen parners working for he counry. Specifically, our conclusion is of direc relevance o he conduc of he moneary policy by he Bangladesh Bank. Keywords: Inflaion, Economic Growh, Granger Causaliy, Srucural Break, Threshold level of Inflaion JEL Classificaion: C13, C22, E31 Respecively Research Economis, Policy Analysis Uni (PAU), Research Deparmen, Bangladesh Bank and Senior Research Associae, Cenre for Policy Dialogue (CPD), Dhaka, Bangladesh. The auhors would like o hank World Bank Insiue (WBI) Residen Economic Adviser a he Bangladesh Bank Prof. Syed M. Ahsan and Md. Ezazul Islam of PAU, Bangladesh Bank for heir helpful suggesions and commens. ii

4 Inflaion and Economic Growh in Bangladesh: Shamim Ahmed & Md. Golam Moraza 1. Inroducion Over he pas few decades, he nexus beween inflaion and economic growh have drawn exensive aenion of macroeconomiss, policy makers and he cenral bankers of boh developed and developing counries. Specifically, he issue ha wheher inflaion is necessary for economic growh or i is harmful generaes a significan debae boh heoreically and empirically. The issue originally evolves from he conroversial noion beween he srucuraliss and he moneariss. 1 In his connecion, Mundell (1965) and Tobin (1965) predic a posiive relaionship beween he rae of inflaion and he rae of capial accumulaion, which in urn, implies a posiive relaionship o he rae of economic growh. 2 They argue ha since money and capial are subsiuable, an increase in he rae of inflaion increases capial accumulaion by shifing porfolio from money o capial, and hereby, simulaing a higher rae of economic growh (Gregorio, 1996). Conversely, Fischer and Modigliani (1978) sugges a negaive and nonlinear relaionship beween he rae of inflaion and economic growh hrough he new growh heory mechanisms (Malla, 1997). They menion ha inflaion resrics economic growh largely by reducing he efficiency of invesmen raher han is level. To dae, alhough he relaionship beween inflaion and economic growh remains conroversial or somewha inconclusive, several empirical sudies confirm he exisence of eiher a posiive or negaive relaionship beween hese wo major macroeconomic variables. Moreover, wih ime a general consensus evolved ha low and sable inflaion promoes economic growh and vice versa (Mubarik, 2005). This furher raises he quesion how low inflaion should be. The answer evidenly depends on he naure and srucure of he economy and varies across counries. In his regard, recenly macroeconomiss have adoped an economeric echnique simply by looking a a nonlinear or srucural break effec which saes ha he impac of inflaion on economic growh could be posiive up o a cerain hreshold level and beyond his level he effec urns o be negaive (Sweidan, 2004). This suppors boh he view of he srucuraliss and he moneariss up o a cerain exen, ha is, low inflaion is helpful for economic growh bu once he economy achieves faser growh hen inflaion is derimenal for he susainabiliy of such growh. The purpose of his paper is o empirically explore he presen relaionship beween inflaion and economic growh in Bangladesh. This has been moivaed by he seminal work of Mallik and Chowdhury (2001) in which hey have performed an economeric 1 The srucuraliss argue ha inflaion is necessary for economic growh, whereas he moneariss argue he opposie, ha is, inflaion as derimenal for economic growh (Mallik and Chowdhury, 2001). 2 Economic growh rae is usually defined as he growh rae of real GDP. 1

5 (i.e., Engle-Granger wo-sep co-inegraion procedure) analysis of he relaionship beween inflaion and economic growh for four Souh Asian counries: Bangladesh, India, Pakisan, and Sri Lanka. 3 Following he works of Khan and Senhadji (2001), Sweidan (2004), and Mubarik (2005), his paper furher explores an ineresing policy issue of how far he inflaion rae is non-derimenal for he economic growh of Bangladesh, or in oher words, wha is he hreshold level of inflaion for he economy. All he empirical analysis of his paper has been conduced using annual daa se on real gross domesic produc (GDP) and consumer price index (CPI) for he period of 1980 o The resuls of he empirical analysis provide guidance for boh domesic policy makers and he developmen parners. However, more migh be learned on inflaion and economic growh using a larger sample (i.e., quarerly daa) over he same ime period and conrol variables. The remainder of his paper is organized as follows: Secion 2 reviews he empirical lieraure on inflaion and economic growh. Secion 3 provides informaion abou he hisorical rends of inflaion and economic growh in Bangladesh. Secion 4 discusses he model and mehodology used o obain he empirical findings repored in his paper. Secion 5 provides daa sources and esimaed resuls on inflaion and economic growh, and finally, secion 6 presens a summary of he main conclusions, limiaions of he paper, and discusses a possible fuure exension. 2. Lieraure Review on Inflaion and Economic Growh Boh in he conex of developed and developing counries, here have been exensive heoreical and empirical research o dae ha aemp o focus on he relaionship beween inflaion and economic growh. This secion presens a brief review. Barro (1995) explores he inflaion economic growh relaionship using a large sample covering more han 100 counries from 1960 o His empirical findings indicae ha here exiss a saisically significan negaive relaionship beween inflaion and economic growh if a cerain number of he counry characerisics (e.g., feriliy rae, educaion, ec.) are held consan. More specifically, an increase he average inflaion by 10 percenage poins per year reduces he growh rae of real per capia GDP by 0.2 o 0.3 percenage poins per year. In oher words, his empirical analysis suggess ha he esimaed relaionship beween inflaion and economic growh is negaive when some reasonable insrumens are considered in he saisical process. Finally, he added ha here is a leas some reason o consider ha higher long-erm inflaion reduces economic growh. Bruno and Easerly (1995) examine he deerminans of economic growh using annual CPI inflaion of 26 counries which experienced inflaion crises during he period beween 1961 and In heir empirical analysis, an inflaion rae of 40 percen and over is considered as he hreshold level for an inflaion crisis. They find inconsisen or somewha inconclusive relaionship beween inflaion and economic growh below his hreshold level when counries wih high inflaion crises are excluded from he sample. In addiion, he empirical analysis suggess ha here exiss 3 A relaed bu no similar sudy has been carried ou by Akharuzzaman (2005) where he examines he dominan facors explaining he inflaionary process in Bangladesh. His resuls suppor ha inflaion is negaively relaed wih real income. 2

6 a emporal negaive relaionship beween inflaion and economic growh beyond his hreshold level. The robusness of he empirical resuls is examined by conrolling for oher facors such as shocks (e.g., erms of rade shocks, poliical crises, and wars). Finally, hey find ha counries recover heir pre-crisis economic growh raes following successful reducion of high inflaion and here is no permanen damage o economic growh due o discree high inflaion crises. Sarel (1995) menions ha inflaion raes were somewha modes in mos counries before he 1970s and afer ha raes sared o be high. Therefore, mos empirical sudies conduced before he 1970s show he evidence of a posiive relaionship beween inflaion and economic growh and a negaive relaionship beween he wo beyond ha ime period due o he severe inflaion hike. Malla (1997) conducs an empirical analysis using a small sample of Asian counries and counries belonging o he Organizaion for Economic Cooperaion and Developmen (OECD) separaely. Afer conrolling for labor and capial inpus, he esimaed resuls sugges ha for he OECD counries here exiss a saisically significan negaive relaionship beween economic growh and inflaion including is firs difference. However, he relaionship is no saisically significan for he developing counries of Asia. The crucial finding of his empirical analysis suggess ha he cross-counry relaionship beween inflaion and long-erm economic growh experiences some fundamenal problems like adjusmen in counry sample and he ime period. Therefore, inconclusive relaionship beween inflaion and economic growh can be drawn from comparing cross counry ime-series regressions wih differen regions and ime periods. Mallik and Chowdhury (2001) examine he shor-run and long-run dynamics of he relaionship beween inflaion and economic growh for four Souh Asian economies: Bangladesh, India, Pakisan, and Sri Lanka. Applying co-inegraion and error correcion models o he annual daa rerieved from he Inernaional Moneary Fund (IMF) Inernaional Financial Saisics (IFS), hey find wo moivaing resuls. Firs, he relaionship beween inflaion and economic growh is posiive and saisically significan for all four counries. Second, he sensiiviy of growh o changes in inflaion raes is smaller han ha of inflaion o changes in growh raes. These resuls have imporan policy implicaions, ha is, alhough moderae inflaion promoes economic growh, faser economic growh absorbs ino inflaion by overheaing he economy. Therefore, hese four counries are on he urning poin of inflaioneconomic growh relaionship. Faria and Carneiro (2001) invesigae he relaionship beween inflaion and economic growh in he conex of Brazil which has been experiencing persisen high inflaion unil recenly. Analyzing a bivariae ime series model (i.e., vecor auoregression) wih annual daa for he period beween 1980 and 1995, hey find ha alhough here exiss a negaive relaionship beween inflaion and economic growh in he shor-run, inflaion does no affec economic growh in he long-run. Their empirical resuls also suppor he superneuraliy concep of money in he long run. 4 This in urn provides 4 Sidrauski (1967) menions ha in an opimal conrol framework if real money balances (M/P) is considered in he uiliy funcion, money is boh neural and superneural. 3

7 empirical evidence agains he view ha inflaion affecs economic growh in he long run. Sweidan (2004) examines wheher he relaionship beween inflaion and economic growh has a srucural breakpoin effec or no for he Jordanian economy from he period beween 1970 and He finds ha his relaion ends o be posiive and significan below an inflaion rae of 2-percen and he srucural breakpoin effec occurs a an inflaion rae equal o 2-percen. Beyond his hreshold level inflaion affecs economic growh negaively. 5 Mubarik (2005) esimaes he hreshold level of inflaion for Pakisan using an annual daa se from he period beween 1973 and He employed he Granger Causaliy es as an applicaion of he hreshold model and finally, he relevan sensiiviy analysis of he model. His esimaion of he hreshold model suggess ha an inflaion rae beyond 9-percen is derimenal for he economic growh of Pakisan. This in urn, suggess ha inflaion rae below he esimaed level of 9-percen is favorable for he economic growh. Moreover, he sensiiviy analysis performed for he robusness of he hreshold model also confirms he same level of hreshold inflaion rae. 3. Hisorical Trends of Inflaion and Economic Growh Bangladesh is he younges counry in he Souh Asian region. Following he launching of a series of comprehensive sabilizaion measures, he economy of Bangladesh mosly resored boh robus economic growh and macro-economic sabiliy in early 1990s from he backdrop of he deep macro-economic crisis of he period since independence (Bhaacharaya, 2004). In paricular, he economy has experienced acceleraed economic growh during he early 1990s in comparison wih he 1980s. However, afer ha period, he economy experienced mos severe exigency saes like increasing inflaionary pressures, deerioraing governmen s budgeary balances and decreasing foreign exchange reserves (Mahmud, 1997). During he firs half of he 1980s he counry experienced a double-digi episode of inflaion while he growh rae of GDP was below 4-percen level as illusraed in Figure 1. The GDP growh rae declined moderaely during he second half of 1980s when inflaion rae gradually decreased o below 8-percen. However, a moderae rae of inflaion and an increasing rae of GDP growh are observed hroughou he 1990s. Throughou he firs half of he 1990s, inflaion rae was, on average, 5.37 percen, while GDP growh rae was 4.06 percen. Alhough inflaion rae increased, on average, o 5.52 percen in he second half of he 1990s, he growh rae of GDP coninued o increase. The increasing rend of inflaion rae during he laer half of 1990s had been correced since he beginning of he new decade afer 1990s and was observed a 4.14 percen, on average, during 2001 o 2005, when growh rae of GDP was, on average, 5.19 percen. 5 Over his period, he average inflaion rae is approximaely 7-percen. 4

8 Figure 1: Five Year Average GDP Growh and Inflaion Raes ( ) rae of inflaion percen GDP growh rae year Source: Daa colleced from BBS (2000, 2001, and 2005). Figure 2 depics he hisorical rends of inflaion rae and real GDP growh rae of Bangladesh during he period of 1981 o Alhough i is unwise o conclude anyhing simply on he basis of a visual inspecion of Figure 2, however, i illusraes more or less an inverse relaionship beween rae of inflaion and GDP growh rae in Bangladesh hroughou his period. Figure 2: Inflaion and Real GDP Growh Raes ( ) inflaion percen GDP growh year Source: Daa colleced from BBS (2000, 2001, and 2005)

9 Moreover, o undersand he hisorical naure of he relaionship beween inflaion and economic growh in Bangladesh more accuraely, he sample covering 1981 o 2005 is grouped ino 7 observaions. 6 Iniially, he range of inflaion is seleced on he basis of he maximum and minimum levels of inflaion from he whole sample. For insance, if inflaion rae is 3-percen or less, i is assigned a level 3. Similarly, he level is 5, if inflaion rae is more han 3-percen bu less han or equal o 5-percen. Subsequenly, wihin his range of inflaion, average GDP growh raes are calculaed agains each linear level of inflaion. For illusraion, wha is he average GDP growh rae when inflaion raes are 3 percen or less during he period from 1981 o 2005 and so on. In his conex, Figure 3 illusraes a posiive relaionship beween inflaion and GDP growh up o he inflaion rae of 7 percen (approximaely) and a negaive relaionship is observed afer ha level of inflaion rae. Figure 3: Average GDP Growh and Linear Level of Inflaion GDP growh average GDP growh Inflaion Source: Daa colleced from BBS (2000, 2001, and 2005). 4. The Model and he Mehodology The paper basically employs wo economeric models o achieve he empirical resuls: he firs one examines he shor-run and long-run relaionships beween real GDP and CPI by applying he Engle-Granger (1987) wo sage co-inegraion procedure and he associaed Error Correcion Model (ECM). 7 In he firs sage, o es for he uni roos of concerned ime series variables, four mos popular echniques have been used: he Dicky-Fuller (DF, 1979) es, he Augmened Dickey-Fuller (ADF, 1981) 6 This par of he analysis follows he approach adoped by Mubarik (2005). 7 The relaionship beween real GDP and CPI in urn implies he relaionship beween inflaion and economic growh. 6

10 es, he Phillips-Perron (PP, 1988) es, and he Kwiakowski-Phillips-Schmid-Shin (KPSS, 1992) es. These ess have been performed in he levels (i.e., log of real GDP and log of CPI) as well as in he firs difference (i.e., economic growh and inflaion rae). If he wo imes series are inegraed of he same order hen he esimaion of he following co-inegraion regression has been considered: GDP = α 11 + β11cpi + ε CPI = α + β GDP + μ (ia) (ib) where, GDP = log of real GDP, CPI = log of CPI, and ε and μ are random error erms or residuals. Furher, economic growh and inflaion rae have been defined as ΔGDP = GDPGR and ΔCPI = INF respecively in he subsequen pars of he 8 paper. In he second sage, he Error Correcion Model (ECM) is employed o see wheher he economy is approaching equilibrium in he long-run or no and he shorrun dynamics of he co-inegraed ime series variables. The ECM is inernally consisen if he wo ime series variables are co-inegraed of he same order or if hey are saionary (Greene, 2003: 654). To deermine he non-saionary propery of hese wo ime series variables boh in he levels and in he firs difference, a firs, he relevan DF, ADF ess have been employed wih and wihou a ime rend. The DF es is based on he following model: Δ χ γ (ii) Z = + ( ρ 1) Z 1 + T + e1 The ADF es is a modificaion over he DF es and lagged values of he dependen variables are added in he esimaion of equaion (ii) which is formed as follows: Δ Z = + ( ρ 1) Z 1 + γt + δδz 1 + e2 χ (iii) Since i is widely believed ha boh DF and ADF ess do no consider he cases of heeroscedasiciy and non-normaliy frequenly revealed in raw daa of economic ime series variables, he PP es for uni roo has been used in he empirical analysis. Moreover, i has an advanage over he ADF es when he concerned ime series has serial correlaion and here is a srucural break. Therefore, he PP es provides robus esimaes over he DF and ADF ess and is based on he following form of equaion: Δ T φ (iv) 2 Z = + ( ρ 1) Z 1 + γ ( ) + ψδz i + e3 The appropriae criical values of he -saisic for he null hypohesis of nonsaionariy are given by MacKinnon (1991). Furher, an alernaive esing procedure, i.e., KPSS es has been performed where he concerned ime series variables are assumed o be rend-saionary under he null hypohesis (Paerson, 2002: 265). The KPSS es sars from he basic local level model: Z = α 1 + β + η + ξ (v) 8 Here, Δ is defined as he firs difference operaor. 7

11 The KPSS es saisic is based on he following lagrangian muliplier (LM) saisic: ( r r 1 T f 2 KPSS = u ) 2 / (vi) 0 where, f is an esimaor of he residual specrum a frequency zero. The appropriae criical values for he LM saisic are given by Kwiakowski-Phillips-Schmid-Shin (1992). In equaions (ii), (iii), and (iv), Δ is defined as he firs difference operaor and, e e are he respecive covariance saionary random error erms. All ess are e1 2, 3 carried ou for boh variables by replacing Z wih GDP and CPI in equaions (ii) (for he DF es), (iii) (for he ADF es), (iv) (for he PP es), and (v) (for he KPSS es). Finally, he DF, ADF, PP, and KPSS uni roo ess have been employed for residuals of equaions (ia) and (ib), i.e., ε and μ. When residuals are found o be inegraed of order zero, I(0), hen i can be concluded ha he wo series, GDP and CPI, are coinegraed and hus a valid and sable long-run relaionship exis beween hem. This also implies he exisence of a sable long-run relaionship beween inflaion and economic growh. Similarly, he Johansen (1988) and Johansen and Juselius (1990) maximum likelihood es procedure is an efficien echnique for esing he coinegraing relaionship beween he concerned ime series variables. This procedure gives wo likelihood raio (LR) ess for he number of co-inegraing vecors, namely, he race es and he maximum eigen value es. Engle and Granger (1987) show ha if wo variables are co-inegraed, i.e., here is a valid long-run relaionship, and hen here exiss a corresponding shor-run relaionship. This is popularly known as he Granger s Represenaion Theorem. Hendry s (1979, 1995) general-o-specific approach has been applied in his case where he model (i.e., ECM) is used in he following form: Δ Δ s q GDP = 10 + ϕ11 j ΔCPI j + ϕ12iδgdp i θ1ε 1 + e4 j= 0 i= 1 ϕ (vii) s q CPI = 20 + ϕ 21 j ΔGDP j + ϕ 22iΔCPI i θ1ε 1 + e4 j = 0 i= 1 ϕ (viii) where, Δ sands for he firs difference operaor, θ 1,θ 2 are he error correcion erms, e4, e5 are he random disurbance erms, and s and q are he number of lag lenghs deermined by he Akaike's informaion crierion (AIC). Here i begins a one and j begins a zero in order for he series o be relaed wihin a srucural ECM (Engle and Yoo, 1991). Finally, 0 θ1, 0 θ 2 should hold for he series o converge o he longrun equilibrium relaion. According o his approach, hree lags of boh he explanaory and dependen variables and one lag of he residual from he coinegraing regression have been included. Subsequenly, he insignifican variables were dropped in order o ge he mos parsimonious model (Hendry, 1979, 1995). I is imporan o menion ha he error correcion erms (i.e., θ 1,θ 2 ) which are he residual series of he co-inegraing vecor normalized for GDP and CPI measure deviaions of hese series from he long-run equilibrium relaions (Mallik and Chowdhury, 2001). 8

12 The second model esimaed in he paper uilizes ha developed by Khan and Senhadji (2001) o esimae he hreshold level of inflaion for Bangladesh above which inflaion affecs economic growh negaively. The equaion o esimae hreshold level of inflaion has been considered in he following condiional form: GDPGR = β + + D( INF K) + U (ix) 0 β1inf β 2 where, K is he hreshold level of inflaion (i.e., he rae of inflaion a which srucural break occurs) and U is he random error erm which represens measuremen error in he explanaory variables. 9 The dummy variable D is defined in he following way: D = 1 if INF > K 0 if INF K The coefficien of he dummy variable ( β 2 ) measures he effec of inflaion rae on he economic growh when i is greaer han he assumed srucural break level (i.e. inflaion is high) and he opposie for he coefficien of inflaion rae ( β 1 ). In he above hreshold model, he sum of he wo coefficiens ( β 1 + β 2 ) represens he annual growh rae of economic growh when inflaion rae is doubled. By esimaing regressions for differen values of K which is chosen in an ascending order (i.e., 1, 2 and so on), he opimal value K is obained by finding he value ha maximizes he R 2 from he respecive regressions. This also implies ha he opimal hreshold level is ha which minimizes he residual sum of squares (RSS). This procedure has become widely acceped in he lieraure on his opic. However, his process is edious since i requires he esimaion of he equaion several imes for differen values of K. 5. Daa and Empirical Evidence The empirical models have used annual daa se on real GDP and CPI for he period of 1980 o 2005 rerieved from he Bangladesh Bureau of Saisics (BBS). 10 For he firs par of he empirical analysis, i.e., he relaionship beween inflaion and economic growh, logs of real GDP ( GDP ) and CPI ( CPI ) have been considered. Furher, economic growh raes (GDPGR) are calculaed from he difference of logs of real GDP and inflaion raes (INF) are calculaed from he difference of logs of CPI (i.e., ΔGDP = GDPGR and ΔCPI = INF respecively) for he second par of he analysis. The summary saisics for GDPGR and INF are repored in Table 1 where he oal number of observaions used in he empirical analysis, means, sandard deviaions, minimum and maximum values of variables during he ime period are given. 9 In he empirical analysis, conrol variables such as invesmen and real gross capial formaion have no been considered due o he unavailabiliy of a consisen ime series daa for Bangladesh. Moreover, money supply (M2) and populaion growh prove saisically insignifican in relaion o economic growh for he sample period of 1981 o 2005, and herefore, have been dropped from he final esimaed equaion. Finally, he number of observaions for all he variables is limied, implying he inabiliy o include many independen variables in order o mainain accepable degrees of freedom. 10 Here, real GDP is calculaed a consan marke prices, while he base year for CPI is , i.e., =

13 Table 1: Summary Saisics of Inflaion and Growh Rae ( ) Variable Observaions Mean Sandard Deviaion Minimum GDPGR (1988) INF (2002) Maximum 6.3 (2004) 15.9 (1982) In Table 2, resuls of he uni roo ess on concerned variables have been repored. The ess for non-saionariy show ha GDPGR is saionary based on DF, ADF, PP, and KPSS ess and in case of INF, DF, PP, and KPSS ess succeed alhough he ADF es fails. Since he PP and KPSS ess are preferable o ADF i can be concluded ha INF is also saionary, I(0). Thus he findings of uni roo ess sugges ha boh he variables GDPGR and INF are inegraed of order zero. Furher, Table 2 shows ha boh GDP and CPI are inegraed of order one based on he DF, ADF, PP, and KPSS ess. Therefore, hey are non saionary, I(1). Variables Wihou rend Table 2: Uni Roo Tess wih DF, ADF, PP, and KPSS DF ADF PP KPSS Decision Wih rend Wihou rend Wih rend Wihou rend Wih rend Wihou rend Wih rend CPI I(1) I(1) I(0) I(1) I(0) I(1) I(1) I(1) I(1) GDP I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) GDPGR I(0) I(0) I(1) I(0) I(1) I(0) I(0) I(0) I(0) INF I(0) I(0) I(0) I(1) I(0) I(0) I(0) I(0) I(0) Noes: 1. means he series in log level, and means significan a 1-percen and 10 percen levels respecively. 2. Lag lengh for ADF ess have been decided on he basis of AIC. 3. Maximum Bandwidh for PP and KPSS ess have been decided on he basis of Newey-Wes (1994). 4. All ess have been performed on he basis of 5-percen significance level using Economeric Views 4 Package. 5. The DF, ADF and PP ess are based on he null hypohesis of uni roos while he KPSS es assumes he null hypohesis of saionariy. In Tables 3 and 4, he esimaed resuls of he relaionship beween CPI and GDP have been repored. They show ha here exiss a long-run and srong inverse relaionship beween CPI and real GDP in Bangladesh which in urn implies a longrun negaive relaionship beween inflaion and economic growh for he counry. The coefficiens are saisically highly significan and negaive for boh regressions (ia) and (ib). Furhermore, Tables 3 and 4 illusrae ha, on average, a 1-percen increase in CPI in Bangladesh leads o a decline in real GDP by 0.19 percen. On he oher hand, on average, a 1-percen increase in he real GDP leads o a decline in CPI rae by 2.38 percen. 10

14 Table 3: Esimaion of he Real GDP Model (ia) Dependen Variable: GDP Mehod: OLS Sample: 1980 o 2005 Variable Coefficien Sandard Error -saisic Probabiliy Consan CPI Time R-squared 0.99 Mean dependen variable Adjused R-squared 0.99 S.D. dependen variable 0.27 S.E. of Regression 0.02 AIC Sum squared residual 0.01 Schwarz crierion Log-likelihood F-saisic Durbin-Wason saisic 0.24 Probabiliy (F-saisic) 0.00 Table 4: Esimaion of he CPI Model (ib) Dependen Variable: CPI Mehod: OLS Sample: 1980 o 2005 Variable Coefficien Sandard Error -saisic Probabiliy Consan GDP Time R-squared 0.98 Mean dependen variable 4.24 Adjused R-squared 0.98 S.D. dependen variable 0.48 S.E. of Regression 0.07 AIC Sum squared residual 0.11 Schwarz crierion Log-likelihood F-saisic Durbin-Wason saisic 0.27 Probabiliy (F-saisic) 0.00 The findings of he esimaed equaions (ia) and (ib) are as follows: (a) here is a linear causaion beween CPI and real GDP in Bangladesh. The esimaed coefficiens are highly saisically significan and negaive implying ha boh CPI and real GDP affec each oher negaively, and (b) inflaion is harmful for economic growh and economic growh helps o reduce inflaion in he counry. Table 5 shows he DF, ADF, PP, and KPSS uni roo ess for residuals of equaions (ia) and (ib), i.e., ε and μ. The resuls sugges ha he residuals are inegraed of order zero, I(0). Therefore, i can be concluded ha he wo series, GDP and CPI, 11

15 are co-inegraed and hus a valid and sable long-run relaionship exis beween hem. Therefore, a sable long-run relaionship beween inflaion and economic growh exiss. Table 5: Uni Roo Tess for he Residuals of (ia) and (ib) Error erm DF ADF PP KPSS Decision ε I(0) I(0) I(0) I(0) I(0) μ I(0) I(0) I(1) I(0) I(0) Noes: 1. Lag lengh for ADF ess have been decided on he basis of AIC. 2. Maximum Bandwidh for PP and KPSS ess have been decided on he basis of Newey- Wes (1994). 3. All ess have been performed on he basis of 5-percen significance level using Economeric Views 4 Package. 4. The DF, ADF and PP ess are based on he null hypohesis of uni roos while he KPSS es assumes he null hypohesis of saionariy. Moreover, he resuls for Johansen maximum likelihood es repored in Table 6 again confirm he rejecion of he null hypohesis of no co-inegraion beween GDP and CPI. In paricular, he compued race, he maximum eigen value saisics and heir corresponding criical values indicae ha he null hypohesis of no co-inegraion (r = 0) can be rejeced under boh of hese ess a boh 5-percen and 1-percen levels of significance. Boh maximum eigen value and race ess indicae one co-inegraing equaion a boh 5-percen and 1-percen levels of significance. This again implies a long-run relaionship beween inflaion and economic growh in Bangladesh. Table 6: Johansen Tes for Co-inegraion Null Hypohesis Alernaive Hypohesis Tes Saisics 5-percen Criical Value Maximum eigen value Tes r = 0 r = r 1 r = percen Criical Value Trace Tes r = 0 r = r 1 r = Conclusion One co-inegraing equaion One co-inegraing equaion Noe: The resuls repored in he above able are based on he assumpions of a consan and a linear rend in he daa wih opimal lag lengh 3. AIC and LR ess have been used o deermine he opimal lag lengh ha makes he residuals whie noise. The second sage of he Engle-Granger procedure comprises of he esimaion of he ECM. Sargan (1984) uses he error correcion mechanism and laer i has been popularized by Engle and Granger who correced ha for disequilibrium. The ECM has several advanages: firs, he ECM incorporaes boh he shor-run and long-run effecs assuming ha he variables are co-inegraed. The second one is ha assuming co-inegraion; all he erms in he model are saionary so ha sandard regression 12

16 echniques are valid (Harris, 1995). The esimaed coefficiens of he error correcion erm (long-run effecs) and he lagged values of he wo series (shor-run effecs) are presened in Table 7. Table 7: The Error Correcion Model Variables Equaion Δ GDP ΔCPI Consan (-0.37) 0.06 (1.51) EC erm (lag = 1) (-3.14) (-2.11) Δ GDP (-1.74) ΔGDP (3.34) 0.64 (0.77) ΔGDP ΔGDP (3.18) Δ CPI ΔCPI (-1.76) (1.67) ΔCPI (2.15) ΔCPI Adjused R-squared DW-saisic Serial Correlaion Funcional Form Normaliy Heeroscedasiciy Noes: 1. Figures in parenheses are -saisics and and indicaes significan a 5- percen and 10 percen levels respecively comparing criical -saisics from sandard -ables. 2. For diagnosics, Godfrey s (1978a, 1978b) LM es for serial correlaion, Ramsey s (1969, 1970) RESET es for funcional form, Whie s (1980) general heeroscedasiciy es for heeroscedasiciy and for normaliy, Jarque-Bera (1980) and Bera-Jarque (1981) ess have been performed. The empirical resuls show he exisence of shor-run and long-run relaionships beween CPI and real GDP in Bangladesh. This also implies shor-run and long-run relaionships beween inflaion and economic growh in he counry. The esimaed coefficiens of he error correcion erms ( θ 1,θ 2 ) are significan a 5-percen level from CPI o real GDP and vice versa wih appropriae (i.e., negaive) signs. Tha means ha in he long-run if he wo series are ou of equilibrium, real GDP will 13

17 adjus o reduce he equilibrium error and vice versa. In oher words, i shows ha 23 percen (error correcion erm -0.23) of he deviaion of he real GDP from is long run equilibrium level is correced each year. On he oher hand, 32 percen (error correcion erm -0.32) of he deviaion of he CPI from is long-run equilibrium level is correced each year. The esimaed resuls in he ECM also show ha shor-run changes in CPI affec real GDP negaively, and vice versa. Therefore, inflaion raes affec economic growh raes negaively, and vice versa. By using ordinary leas squares (OLS), Table 8 gives he exac value of he hreshold inflaion level and also shows he impac of ha inflaion level on economic growh by esimaing equaion (vi). The esimaed value of R 2 is aken ino consideraion by esimaing equaion (vi) for he hreshold level of inflaion considering K = 1 o K = 11. However, considering he value of R 2, he esimaed resuls have been repored in he Table 7 for K values ranging from 4-percen o 8-percen. For he period beween 1981 and 2005, a significan and consisen resul can be found if only inflaion, lagged by hree periods, (i.e., lag = 3) is considered o esimae he exac hreshold level of inflaion. Therefore inflaion is kep a lag hree in he esimae. As reviewed in he previous secion, in his approach, he hreshold level is one ha maximizes he value of R 2. Table 8: Economic Growh Model wih he Srucural Break Term Dependen Variable: GDPGR K Variable Coefficien Sandard Error 4-percen INF D(INF-K) C percen 6-percen 7-percen 8-percen INF D(INF-K) C INF D(INF-K) C INF D(INF-K) C INF D(INF-K) C saisic Probabiliy R-squared From he esimaed resuls, i is observable ha a low hreshold inflaion levels (K< 6) here is a saisically insignifican relaionship (a 5-percen level) beween he dummy of hreshold level of inflaion and economic growh. As K increases saring from 6-percen, a saisically significan relaionship (a 5-percen level) is observed beween economic growh and he dummy of hreshold level of inflaion which coninues up o 7-percen inflaion rae. However, in he esimaion process, he hreshold level of inflaion is observed a 6-percen level where he value of R 2 is maximized i.e. RSS is minimized. While inflaion below his hreshold level has no 14

18 significan effec on economic growh (i.e., saisically insignifican a 5-percen level), inflaion raes above i has a significan negaive effec on economic growh. Therefore, he empirical analysis suggess ha if inflaion rae is above 6-percen, hen he economic growh performance of Bangladesh migh experience a jeopardized siuaion. 6. Conclusion This paper empirically explores he presen relaionship beween inflaion and economic growh in he conex of Bangladesh. An assessmen of he empirical evidence has been acquired hrough he co-inegraion and error correcion models. Furher, he paper explores an ineresing policy issue of wha is he hreshold level of inflaion for he economy. The empirical evidence demonsraes ha here exiss a saisically significan long-run negaive relaionship beween inflaion and economic growh for he counry as indicaed by a saisically significan long-run negaive relaionship beween CPI and real GDP. This resul is more or less consisen wih he predicions of Mallik and Chowdhury (2001). Paricularly, hey have menioned ha Bangladesh was already on he urning poin (i.e., from posiive o negaive) of inflaion-economic growh relaionship in he lae 1990s. In addiion, he esimaed hreshold model suggess 6-percen hreshold level (i.e., srucural break poin) of inflaion above which inflaion adversely affecs economic growh. These resuls have imporan policy implicaions for boh domesic policy makers and he developmen parners. Firs, aking ino consideraion ha he inflaion rae is no indexed in he wages and salaries, inflaion will lead o a decrease in he purchasing power and an increase in he cos of living. Second, given ha he counry frequenly has o balance he credi requiremens by he privae and public secor agains boh inflaionary and balance of paymens pressures, i is no always possible for he moneary auhoriy o increase (or adjus) he nominal ineres rae above he expeced (or acual) inflaion rae hrough conracionary moneary policy. 11 In his regard, he moneary auhoriy can hink of an alernaive way by working on he expecaions channel o reduce inflaion. This requires credibiliy of he moneary auhoriy in following hrough is moneary program as communicaed in advance o he sakeholders. Some caveas are in order. For example, in he conex of Bangladesh, he empirical resuls provided in his paper do no address he following imporan issues: 1. I does no esimae how he economic growh rae will behave as he rae of inflaion rises, bu remaining conained wihin he hreshold level. 11 The New Keynesian model suggess ha o reduce inflaion, he Cenral Bank should lean agains he wind. Tha is, o increase he nominal ineres rae above he expeced (or acual) inflaion rae hrough he conracionary moneary policy. The model is basically summarized by he following wo equaions: 1 x E x i E π + u [ ] = σ = βe π + + κx + e x = oupu gap, π = inflaion rae, and i = π 1 where, nominal ineres rae. 15

19 2. Does higher inflaion lead o greaer inflaion uncerainy? In paricular, wha is he relaionship beween inflaion and inflaion uncerainy? 3. Wheher inflaion uncerainy adversely affecs economic growh. 4. Wha are he deerminans of high inflaionary pressure? Fuure research should exend in he above direcions in order o derive firm policy relevan conclusions. References Akharuzzaman, Md. Inflaion in he Open Economy: An Applicaion of he Error Correcion Approach o he Recen Experience in Bangladesh, Working Paper Series, WP 0602 (2005), Policy Analysis Uni (PAU), Research Deparmen, Bangladesh Bank. Bangladesh Bureau of Saisics. Naional Accouns Saisics of Bangladesh (Revised Esimaes, o ), Srenghening Naional Accouns and Povery Monioring Projec (SNAPMP), Naional Accouning Wing (NAW), BBS, Minisry of Planning: Saisics Division (2000).. Naional Accouns Saisics (Gross Domesic Produc, ), NAW, BBS, Minisry of Planning: Saisics Division (2001).. Naional Accouns Saisics (Provisional Esimaes of GDP, and Final Esimaes of GDP, ), SNAPMP, NAW, BBS, Minisry of Planning: Planning Division (2005). Barro, R. J. Inflaion and Economic Growh, Naional Bureau of Economic Research (NBER) Working Paper No (Ocober 1995). Bera, A. K. and C. M. Jarque. An Efficien Large-Sample Tes for Normaliy of Observaions and Regression Residuals, Ausralian Naional Universiy Working Papers in Economerics, Vol. 40 (1981), Canberra. Bhaacharaya, D. Bangladesh Economy: Macroeconomic Performance, mimeo, Cenre for Policy Dialogue, Dhaka (2004). Available a (access dae: 1, December 2005). Bruno, M. and W. Easerly. Inflaion Crises and Long-Run Growh, World Bank Policy Research Working Paper No (1995). 16

20 Culver, E. S. and H. D. Papell. Pifalls and Opporuniies: Wha Macroeconomics Should Know abou Uni Roos, in O. J. Blanchard and S. Fischer, eds., NBER Macroeconomics Annual 1991, Cambridge: MIT Press (1997). Dickey, D. A. and W. A. Fuller. Disribuion of he Esimaors for Auoregressive Time Series wih a Uni Roo, Journal of he American Saisical Associaion, Vol. 74 (1979), pp Likelihood Raio Saisics for Auoregressive Time Series wih a Uni Roo, Economerica, Vol. 49 (1981), pp Engle, R. F. and C. W. J. Granger. Co-inegraion and Error Correcion: Represenaion, Esimaion and Tesing, Economerica, Vol. 55 (1987), pp Engle, R. F. and B. S. Yoo. Co-inegraed Economic Time Series: An Overview wih New Resuls, in R. F. Engle and C. W. J. Granger, eds., Long-Run Economic Relaionships, Oxford: Oxford Universiy Press (1991), pp Faria, J. R. and F. G. Carneiro. Does High Inflaion Affec Growh in he Long and Shorrun?, Journal of Applied Economics, Vol. IV, No. 1 (2001), pp Fischer, S. and F. Modigliani. Towards and Undersanding of he Real Effecs and Coss of Inflaion, Welwirschafliches Archiv, 114 (1978), pp Godfrey, L. G. (a). Tesing Agains General Auoregressive and Moving Average Error Models When he Regressors Include Lagged Dependen Variables, Economerica, Vol. 46 (1978), pp (b). Tesing for Higher Order Serial Correlaion in Regression Equaions When he Regressors Include Lagged Dependen Variables, Economerica, Vol. 46 (1978), pp Gregorio, Jose de. Inflaion, Growh, and Cenral Banks: Theory and Evidence, World Bank Policy Research Working Paper 1575, Policy Research Deparmen, Macroeconomics and Growh Division (February 1996). Greene, W. H. Economeric Analysis. 5 h ed. New Jersey: Prenice-Hall (2003). Harris, R. I. D. Using Co-inegraion Analysis in Economeric Modeling, London: Prenice Hall, Harveser Wheasheaf (1995). Hendry, D. F. Producive Failure Economeric Modelling in Macroeconomics: The Transmission Demand for Money, in Economic Modeling: Curren Issues and Problems in Macroeconomic Modeling in he UK and he USA, (ed.), London: P. Ormerod (1979).. Dynamic Economerics, London: Oxford Universiy Press (1995), pp Jarque, C. M. and A. K. Bera. Efficien Tess for Normaliy, Homoscedasiciy and Serial Independence of Regression Residuals, Economics Leers, Vol. 6 (1980), pp Johansen, S. Saisical Analysis of Co-inegraion Vecors, Journal of Economic Dynamics and Conrol, Vol. 12 (1988), pp

21 Johansen, S. and K. Juselius. Maximum Likelihood Esimaion and Inference on Coinegraion wih he Applicaion o he Demand for Money, Oxford Bullein of Economics and Saisics, Vol. 52 (1990), pp Kwiakowski, D., P. Phillips, P. Schmid and Y. Shin. Tesing he Null Hypohesis of Saionariy Agains he Alernaive of a Uni Roo, Journal of Economerics, Vol. 54 (1992), pp Khan, M. S. and A. S. Senhadji. Threshold Effecs in he Relaionship beween Inflaion and Growh, IMF Saff Papers, Vol. 48, No. 1 (2001). MacKinnon, J. Criical Values for Co-inegraion Tess, In R. F. Engel and C. W. J. Granger, eds., Long Run Economic Relaionships: Readings in Co-inegraion, Oxford: Oxford Universiy Press (1991). Mahmud, W. Growh or Sagnaion?: A Review of Bangladesh's Developmen 1996, Macroeconomic Updae, Cenre for Policy Dialogue, Dhaka: Universiy Press Limied (1997). Malla, S. Inflaion and Economic Growh: Evidence from a Growh Equaion, mimeo, Deparmen of Economics, Universiy of Hawai I a Monoa, Honolulu (1997). Available a (access dae: 25, November 2005). Mallik, G. and A. Chowdhury. Inflaion and Economic Growh: Evidence from Souh Asian Counries, Asian Pacific Developmen Journal, Vol. 8, No.1. (2001), pp Mubarik, Y. A. Inflaion and Growh: An Esimae of he Threshold Level of Inflaion in Pakisan, Sae Bank of Pakisan Research Bullein, Vol.1, No. 1-2 (2005), pp Mundell, R. Growh, Sabiliy and Inflaionary Finance, Journal of Poliical Economy, 73 (1965), pp Newey, W. and K. Wes. Auomaic Lag Selecion in Covariance Marix Esimaion, Review of Economic Sudies, Vol. 61 (1994), pp Paerson, K. An Inroducion o Applied Economerics: A Time Series Approach. New York: Palgrave (2002). Phillips, P. C. B. and P. Perron. Tesing for a Uni Roo in Time Series Regression, Biomerika, Vol. 32 (1998), pp Ramsey, J. B. Tes for Specificaion Errors in Classical Linear Leas Square Regression Analysis, Journal of he Royal Saisical Sociey B (1969), pp Models, Specificaion Error and Inference: A Discussion of Some Problems in Economeric Mehodology, Bullein of he Oxford Insiue of Economics and Saisics, Vol. 32 (1970), pp Sarel, M. Nonlinear Effecs of Inflaion on Economic Growh, IMF Working Paper WP/95/56, Washingon (May 1995). Sragan, J. D. Wages and Prices in he Unied Kingdom: A Sudy in Economic Mehodology, Originally published in 1964 and reproduced in K. F. Wallis and D. F. Hendry, eds., Quanaive Economics and Economeric Analysis, Oxford: Basil Blakewell (1984). 18

22 Sidrauski, M. Raional Choice and Paerns of Growh in a Moneary Economy, American Economic Review 57 (1967), pp Sweidan, O. D. Does Inflaion Harm Economic Growh in Jordan? An Economeric Analysis for he Period , Inernaional Journal of Applied Economerics and Quaniaive Sudies, Vol.1-2 (2004), pp Tobin, J. Money and Economic Growh, Economerica, 33 (1965), pp Whie, H. A Heeroscedasiciy Consisen Variance Marix Esimaor and a Direc Tes of Heeroscedasiciy. Economerica, Vol. 48 (1980), pp