BATH ECONOMICS RESEARCH PAPERS

Transcription

1 Revisiing he Dynamic Effecs of Oil Price Shocks on Small Developing Economies Imran H. Shah, Carlos Diaz Vela, Yuan Wang No. 65 /17 BATH ECONOMICS RESEARCH PAPERS Deparmen of Economics

2 Revisiing he Dynamic Effecs of Oil Price Shocks on Small Developing Economies 17 March Imran H. Shah (Firs and corresponding auhor), Deparmen of Economics, Universiy of Bah, BA2 OJH, Unied Kingdom. Phone: Carlos Diaz Vela (Co-Auhor), Universiy of Leiceser, UK. 3. Yuan Wang (Co-Auhor), Universiy of Souhampon, UK. ABSTRACT This paper examines he dynamic effecs of oil price, aggregae demand and aggregae supply shocks on oupu and inflaion in four small developing economies using a srucural VAR model. For all counries, despie finding he expeced response of oupu o oil price shocks, an upward causal effec of oil price innovaions on he domesic price level is esablished which adversely accompanies he growh simulaing effecs in oil-exporing counries. This paper also finds asymmeric effecs of oil price changes on macroeconomic variables in all sample counries. Finally, our empirical resuls find wo furher hings: firsly, ha for Malaysia, Pakisan and Thailand, nominal demand and supply shocks are he main sources of flucuaions in inflaion and oupu respecively, whereas for Indonesia he converse holds and secondly ha whils he 1998 recession was largely induced by only supply and demand shocks he recession of could poenially be explained by oil price changes. Key words: Macroeconomic flucuaions, Oil price, Srucural VAR models, Asian developing economies JEL Classificaion Numbers: C32, E31, E32, F41, O40, O53

3 1. INTRODUCTION In he exising lieraure, he relaionship beween changes of oil price (OP) and macroeconomic flucuaions has been inensively discussed (e.g., Abeysinghe, 2001; Bjornland, 2000; Hamilon, 1983, 1996, 2003, 2009 and 2011; Kilian, 2008 and 2009; Kilian and Vigfusson, 2013; Mork, 1989). A large amoun of empirical evidence suppors he exisence of a causal relaionship from OP changes o oupu flucuaions (Cunado and Perez de Gracia, 2003; Jimenez-Rodriguez and Sanchez, 2005) and also in he variaions of domesic aggregae price level (Barsky and Kilian, 2004; Elder and Serleis, 2010; Roemberg and Woodford, 1996). The seminal conribuion in his area is credied o Hamilon (1983), he empirical work shows ha he OP shocks conribue o some recessions in he US before 1972, bu no necessarily lead a recession. In alignmen wih he previous finding, Hamilon (1996) claims a non-linear relaionship beween OP and oupu. More precisely, OP increases (wheher he OP has exceeded he maximum OP in recen years) are believed o be more imporan han OP decreases in predicing oupu growh, as suggesed by Hamilon. Elder and Serleis (2010) find ha he volailiy of he OP has saisically significan negaive effecs on oupu: exacerbaes he negaive impac o a negaive OP shock and dampens he impac o a posiive OP shock. The recen work of Kilian and Vigfusson (2013) saes ha he OP is no helpful for ou-of-sample forecasing in he asymmery nonlinear models, bu useful in symmeric nonlinear models. I is possible o forecas he recession by using he laer. Convenional wisdom suggess ha he correlaion beween OP shocks and macroeconomic flucuaions comes from a se of fundamenal shocks, which hi all secors of he economy. A large number of sudies has focused on idenifying he source of ha shocks. The seminal ideas come from Hamilon (1983) and Kilian (2009), who focus on he demand and supply side respecively. Boh auhors provide exraordinary explanaions and rigorous proofs. Hamilon believes ha oil supply shocks, usually caused by a shorfall of OPEC oil producion due o poliical evens in he Middle Eas, has a major role in explaining he dynamics of OP and real GDP movemens. However, Kilian (2009) believes ha oil demand and speculaive OP shocks linked wih global economic aciviy have being he major driving force of oil price flucuaions since The variaions of OP are expeced o have very differen effecs on oil-imporing and oil-exporing counries. Brown e al. (2004) show ha increased OP has a similar effec o a ax, which is colleced by oil-exporing counries from oil-imporing counries. Mos of he 2

4 empirical sudies mainly relaed o developed economies have confirmed ha OP changes have a negaive effec on oupu growh in oil-imporing counries (e.g., Jimenez-Rodriguez and Sanchez, 2005; Schneider, 2004; Schuber and Turnovsky, 2011). Increases in OP and he resulan insabiliy affec he economy hrough higher inpu coss, reallocaion of resources, decreases in income and depreciaion of currency. As a resul, economic performance is depressed while inflaion and unemploymen are raised. A sudden increase in he OP causes an exogenous inflaionary shock as high OP pus pressure on he general price level, and leads o higher ineres raes. The former is very likely lead o a recession. Oher sudies (e.g., Bjornland, 2000) indicae ha an increase in OP is associaed wih higher growh in ne oil-exporing counries hrough increased sae revenue, which leads o higher naional income and currency appreciaion. Bjornland (2009) saes ha high OP also affecs oil-producing counries hrough negaive rade effecs by rising inflaion and since oilimporing counries suffer from oil induced recession and herefore demand fewer expors from oil-exporing counries. For hose oil-exporing counries wih a large expor secor, he negaive rade effec may indeed off-se he posiive wealh effec, which leaves he ne effec ambiguous. The recen experience in Asian counries somehow forms he hypohesis of nonnegligible effecs of OP shocks on economic flucuaion under some circumsances. Cunado and Perez de Gracia (2005) invesigae poenial impacs of OP on economic aciviy and consumer price indexes for six Asian counries and hey find ha resuls are sensiive o he choice of he OP (domesic or inernaional). The shor-run effecs of OP on economic growh and inflaion are saisically significan. Spikes in OPs prior o he global crisis led o high inflaion raes in mos Souh and Souh Eas Asian counries evidenced by double-digi figures. The inflaionary pressures also induced large budge deficis and balance of paymen concerns. Lescaroux and Mignon (2009) and Tang e. al. (2010) boh find negaive effecs of OP on macroeconomic variables in China. However, in he recen work of Schuber and Turnovsky (2011), he evidence shows ha recen OP shocks have a empered effec on economic aciviy on developing counries compared wih hose in 1970s and 1980s. 1 One obvious explanaion would be he fall in boh proporion and influence of oil in he economy due o more credible policy response, more flexible labour marke and appearance of oil subsiues. 1 See Blanchard and Gali (2008) for a deailed review. 3

5 Mork (1989), Lee e al. (1995) and Hamilon (1996) inroduced asymmeric specificaions of he OP o reconsruc he relaionship beween OP and economic performance. I has been assered in he lieraure ha he relaionship beween OP shocks and macroeconomic variables is non-linear and many sudies have suggesed he possibiliy of an asymmeric impac of OP shocks on macroeconomic variables (Mork, 1989; Lee and Ni, 2002; Jimenez-Rodriguez and Sanchez, 2005; Hamilon, 2009 and 2011; An e al., 2014, Kilian and Vigfusson, 2011). According o hese sudies, he negaive effecs of higher OPs are greaer han he posiive effecs of lower OPs in oil-imporing counries. On he oher hand, Moshiri (2015) examines he asymmeric effec of OP shocks on macroeconomic performance in he nine major oil-exporing counries using a VAR model. He finds ha he negaive OP shocks have damaging impacs on oupu of oil-exporing counries, however posiive OP shocks do no simulae oupu. The main conribuion of his paper is o exend he exising empirical lieraure in wo direcions. Firsly, we invesigae he linear effecs of OP shocks, in addiion o nominal AD and AS shocks, on oupu and inflaion in four Asian developing counries: Indonesia, Malaysia, Pakisan and Thailand. The reason behind our invesigaion of hese counries is ha here is lile previous evidence on he effecs. Secondly, we invesigae he asymmeric effecs of OP by using a non-linear mehod, namely he asymmeric specificaion. The conjecure is ha posiive oil price shocks could have larger effecs on growh and ha negaive oil price shocks do no subsanially affec growh for oil-imporing counries (and vice versa for oil-exporers). The sudy uses a srucural VAR approach, which imposes boh shor-run and long-run resricions o idenify differen srucural shocks and o carry ou boh linear and non-linear models. Indonesia is he larges oil producer in Souheas Asia and is also a major oil-exporer during he period analysed due o is considerable oil reserves. According o he World Bank, Indonesian expors accoun for abou 40% of GDP. Malaysia has he hird highes oil reserves of he Souheas Asian counries bu is ne oil-expors are very small due o he small gap beween domesic producion and demand. Thailand produces some oil domesically, however i is sill a significan oil-imporing counry due o he large domesic demand of oil. In Thailand, wo hirds of oil demand is impored from abroad, which accouns for a large proporion of GDP. Pakisan is also an oil-imporing counry since i has limied domesic oil reserves and relies heavily on impors. However, Pakisan demands he smalles amoun of oil among he four sample economies. Oil-imporing developing economies are generally 4

6 considered highly vulnerable o exernal shocks, and prominen among hese is volailiy in he OP because an OP increase ransfers income from counries wih a higher propensiy o consume o counries wih lower propensiy o consume. Given he above facs abou producion and demand, his sudy will consider Indonesia and Malaysia as oil-exporing counries, and Pakisan and Thailand as oil-imporing counries. Alhough Indonesia and Malaysia are ne oil-exporers, due o he exhausion of oil fields and he lack of invesmen in exploraion of oil, he economy experienced sagnan oil producion. Consequenly, he oil marke is relaively igh due o he small gap beween domesic producion and oil demand. I is ineresing o invesigae he impac of he OP on counries which are in he process of ransforming hemselves from ne-exporers o neimporers. For all sample counries, governmens provide subsidies on he OP in order o reduce he adverse effec of OP shocks on real economic aciviies. Consequenly, when he inernaional OP rose from 50 dollars per barrel in 2007 o 140 dollars per barrel in 2008 a significan budge defici was induced. According o he World Bank (2013) 2, he raios of governmen budge defici o GDP increased from -0.08% o -1.6% in Indonesia, -4.6% o - 6.7% in Malaysia, -3.9% o -7.2% in Pakisan and -1.3% o -4.8% in Thailand afer he 2008 recession. In he pas hree decades, hese counries have experienced an increased demand for energy, paricularly for oil following he deepening of economic developmen. There has recenly been a period of negaive growh which was dominaed by he global financial crisis and also wih he adverse effec of an OP shock. Idenifying and undersanding he effecs of various shocks on macroeconomic flucuaions in he sample counries, paricularly he effecs of OP shocks, could provide some policy recommendaions for regional co-operaion in he Asian-Pacific area. This could help o minimize negaive influences from global economic flucuaions and achieve macroeconomic sabiliy for susainable growh and developmen. The res of he paper is organised as follows. Secion 2 inroduces a heoreical model wih separaed OP shocks in addiion o AD and AS shocks. The empirical mehodology and esimaion resuls are repored in Secion 3. Finally, a few concluding remarks are provided in Secion THEORETICAL MODEL 2 hp://sieresources.worldbank.org/dec/resources/ / / /oil_price_volailiy.pdf 5

7 The heoreical framework in his paper is a modified AD and AS model. The effecs of differen srucural shocks on macroeconomic variables have been modelled explicily. The Lucas supply curve (Lucas, 1972 and 1973) wih raional expecaions can be defined as: y y [ p E ( p )] s 1 1 (1) s where AS y, is a funcion of naural rae of oupu y and he difference beween acual p 1 domesic price level and is expecaion given all available pas informaion. Taking he expecaion condiional on ime -1 and rearranging Eq. (1) gives: y E ( y ) [ p E ( p )] s (2) where represens produciviy shock, which can be furher decomposed ino a supply (defined as domesic supply) shock and an OP shock (see Bjornland, 2000). y E ( y ) [ p E ( p )] s s op (3) Eq. (3) describes he AS curve where oupu increases as a resul of an unprediced increase op s in price levels, an OP shock and a posiive realizaion of he AS shock. High OPs are considered as echnology shocks, which reduce oupu level hrough increasing producion cos in oil-imporing counries. In anoher words, he oil price affecs he supply side of he economy by increasing he cos of inpus and necessiaing a rearrangemen of resources, hus leading o lower GDP. 3 Oil price hikes may add o inflaionary pressures and reduce real incomes and, as a resul, consumer expendiure will be compressed. Furhermore, real oupu may also be affeced in he face of weaker domesic demand and reduced company's profiabiliy. Research s finds ha OP affecs oupu and price levels, and he lower level of demand can be couneraced hrough he corresponding moneary policies implemened by he cenral banks (Hamilon, 1996, Hamilon, 2003 and Hooker, 2002). High OPs affec he economy of oil-imporing counries hrough increased marginal coss and inflaion. I is herefore expeced ha 0 for oil-imporing counries. In conras, oil-exporing counries will respond o he same shock posiively ( 0 ) due o an increase in naional income 3 Brown e al. (2004) argues ha rising OPs decrease purchasing power and consumer demand in oil-imporing economies, and ha he opposie should be expeced for oil-exporing naions. 6

8 hrough greaer oil expor revenue. This is especially applicable o hose oil-exporing counries where he oil secor is large compared wih he res of he economy. The demand curve can be expressed as: y m p op d (4) d where AD, y, is a funcion of lieral money m, domesic price level p and oil price op. Similarly as for he supply side, aking expecaions condiional on ime 1 and rearranging Eq. (4) gives, y E ( y ) [ p E ( p )] d d op (5) Eq. (5) explains ha AD equals is expeced value given he informaion available a he end of period -1, plus he effec of an OP shock op and nominal demand (defined as a price shock) shock d. Oil price flucuaions can also affec he economy hrough he demand side via he income effec. Spikes in he OP will shif income from oil-imporing counries o ne oil-exporing counries. Therefore, in each caegory we would expec gamma γ o be less han 0 and greaer han 0 respecively. The economy is in equilibrium when, y y y s d (6) Hence we have, 1 1 p E ( p ) s d op y E ( y ) s d op 1 1 (7) (8) Following Bjornland (2000), we assume ha world OP can only be affeced by shocks o oil demand and oil supply, while oher facors (such as poliical evens) are considered as exogenous o he OP. 4 Hence, 4 In paricularly, our sample counries are small economies. Hence, his is a reasonable assumpion. 7

9 op op w w op 1 (9) The early sudies included Blanchard and Gali, 2010; Bjornland, 2000; Hamilon, 1983, 2003; Jimenez-Rodriguez and Sanchez, 2005; Schmiz el al, 2007; Tang, 2010 who regard OP changes are as exogenous o he conemporaneous value of macroeconomic variables. Furhermore, hey prove ha oil shocks affec macroeconomic variables whereas changes in hese variables canno affec OP which is exogenously deermined by he world s demand and supply shocks. The reason behind his is ha OP has been dominaed by poliical evens like he OPEC embargo in 1973, he Iranian revaluaion in , he Iran-Iraq War in , he Gulf War in , and increasing demand confroning declining world producion in Eqs. (7) - (9) give us he srucural form model in he paper. Each srucural shock (OP op d s shock,, AD shock, and AS shock, ) is a whie noise and hey are assumed o be uncorrelaed wih each oher. In he shor run, OP, AS and AD shocks affec he oupu level due o nominal and real inflexibiliies as exhibied in Eq. (8). In alignmen wih Blanchard and Quah (1989), we assume AS shocks have permanen effec on oupu level, while AD shocks only affec oupu level in he shor run. The OP shocks are assumed o be exogenous, which is consisen wih he recenly work of Kilian and Vigfusson (2013). 3. EMPIRICAL METHODOLOGY AND RESULTS 3.1 Daa Descripion In his secion we invesigae he response of differen ypes of shocks ino four Asian counries. The daa used in his sudy is real OP, real GDP and consumer price index (CPI) for 4 sample economies: Indonesia, Malaysia, Pakisan 5 and Thailand. 6 For all counries, 5 For Pakisan, quarerly daa series for GDP are no publish or are no available for a longer period of ime. The indusrial producion index is used as a proxy of he real GDP. The quarerly series for he indusrial producion is downloaded from IMF, IFS, Alhough, indusrial producion daa for Pakisan is available before 1990 bu in order o be consisen wih oher counries we consider i from Daa are downloaded from he Inernaional Moneary Fund (IMF), Inernaional Financial Saisics (IFS), Ediion: July The real GDP series are based on auhors calculaion from he GROSS DOMESTIC PRODUCT (GDP) (Unis: Naional Currency; Scale: Billions) and GDP DEFLATOR (2005=100) for wo counries excep Indonesia, where he consumers price index, CPI, has been used as he deflaor.. The domesic nominal OP series is calculaed from PETROLEUM AVERAGE CRUDE PRICE (Unis: US Dollars per Barrel) convered o each counry s naional currency. For Malaysia and Thailand, he real OP is compued from deflaed by implici consumer price index, while for Indonesia and Pakisan, where he CPI has been used as he 8

10 quarerly GDP daa is no available for an exended period of ime and herefore, he sample sars a he earlies available dae. The ime-spans differ across counries depending on he availabiliy of daa: Indonesia (1990q1 o 2015q4), Malaysia (1991q1-2015q4), Pakisan (1990q1-2015q4) and Thailand (1993q1-2015q4). The daa are ransformed by aking logarihms. 3.2 Empirical Model and Esimaion Resuls As shown in Appendix A, Table A1, he uni roo ess (boh Augmened Dickey Fuller and Phillips Perron ess) indicae ha none of hese hree series are saionary, so ha all of hem are differenced before esimaion. Before proceeding furher, a few pre-esimaion ess are conduced. As exhibied in Appendix A, Table A2, we use he approach inroduced by Zivo and Andrews (1992) o es he null of uni roo agains srucural break-saionary alernaive hypohesis. Furhermore, we es he OP growh, inflaion and GDP growh daa wih several uni roo ess wih srucural breaks in essenially rend-saionary series, namely, ERS-P T Ellio-Rohenberg-Sock, NP-MZα Ng-Perron, SKP-MSB Silvesre-Kim- Perron, SKP-MZ T Silvesre-Kim-Perron and PP- Zα Phillips-Perron 7.The resuls indicae ha all of hese hree series are saionary a firs difference for each counry (see Appendix A, Table A2). The resuls are consisen wih he convenional uni roo es wihou srucural breaks. For all counries, he break period for each series is repored in able Appendix A, Table A2. The Johansen es of coinegraion is performed nex o he level series wih appropriae assumpions on rends and lags o check wheher he variables are coinegraed in each counry (see Appendix Table A3). Generally speaking, here is no coinegraion evidence among hree series in any counry. 8 Noice, hen, ha we are modelling he rae of growh of GPD and OP as well as inflaion. A reduced-form VAR (k) model is fied o he saionary daa where he a priori assumpion of weakly exogenous OP is imposed. Therefore, denoing he lower case names of he variables as he log-ransformed values of oil prices (op), GDP (y) and CPI (p), he model used for esimaion is deflaor. The CPI series are given as follows: CPI:17 CAPITAL CITIES (Indonesia), CPI PENINSULAR MALAYSIA (Malaysia), CPI:12MAJOR CITIES ALL INC. (Pakisan), CPI: URBAN (Thailand). 7 Ng-Perron (2001) proposed four es saisics ha are based upon he GLS-derended daa. 8 Boh he race and maximum eigenvalues es saisics indicae no coinegraion a he 0.05 level for Indonesia, Malaysia and Pakisan, while a he 0.01 level for Thailand. 9

11 k op ' 1 x1 11 j j j 1 op a a op e k k k s ' 2 x2 21 j j 22 j j 23 j j j 1 j 1 j 1 y a a op a y a p e k k k d ' 3 x3 31 j j 32 j j 33 j j j 1 j 1 j 1 p a a op a y a p e (10) where (j=1,2,3) is a vecor conaining deerminisic variables used o model variable j. This vecor conains a consan ha accouns for he possibiliy of he presence of a deerminisic ime rend in he growh of he variables and a se of inervenion variables o accoun for he srucural breaks suggesed by he saionariy ess commened above 9. More deailed informaion regarding he esimaion resuls can be obained from he auhors on reques. The exogeneiy propery of OP growh is manifesed in he firs equaion of sysem (10) as GDP growh and inflaion do no affec OP, which is driven only by is pas dynamic. To validae he exogeneiy assumpion of he OP, we perform a Granger causaliy es o examine wheher he lagged values of growh and inflaion have explanaory power on OP growh and he resuls are repored in Appendix A4. The resuls show ha OP is no affeced by real GDP for all counries. The finding is consisen wih previous sudies such as Bjornland (2000) and Tang (2010). Furhermore, inflaion does no Granger cause OP for Indonesia and Pakisan bu for Malaysia and Thailand he exogeneiy condiion for OP is valid only a 1% level of significance. The model can be wrien in compac form as = + +, (11) e e op s d where = (,, ) and =,, ~ (, ), where is a posiive definie marix conaining he variance and covariances of he errors of he reduced form model. Esimaion of his model is performed using he JMulTi sofware 10 and EViews. Summarized esimaion resuls are shown in Table The firs wo rows indicae he number of effecive observaions used in esimaion and he auoregressive order used in each model. For all counries he opimal lag seleced by he Bayesian Informaion Crieria indicaes a large value of 2. The nex wo rows indicaes he value of he mulivariae Box-Pierce pormaneau es e 9 The resuls are no repored bu up o reques. 10 Free sofware available for download a hp:// accompanying Lükepohl (2006). 11 Esimaion has been performed also using annual daa for Pakisan and resuls seem o be robus. 10

12 for auocorrelaion along wih he corresponding p-values, indicaing ha here is no significan correlaion lef in each model. [TABLE 1] Table 1: Model Esimaion Resuls Indonesia Malaysia Pakisan Thailand N K Q p-val (Q) M-JB p-val (JB) e e e e-10 Llik Nex, he mulivariae Jarque-Bera es indicaes deparure from he assumpion of normaliy of he residuals. However, given he good fi of he model, and ha here are no significan ouliers lef unmodelled, we consider he esimaes saisfacory. The maximum eigenvalue of he auoregressive polynomial ( ) indicaes no problem of persisence or coinegraion as esed iniially using Johansen s coinegraion ess (see Appendix B). Finally, he value of he likelihood funcion, he deerminan of he covariance marix and he Akaike and Bayesian Informaion Crieria are also repored. Esimaes have been checked for robusness indicaing sabiliy of he esimaes. The impulse-response analysis is performed using he Wold decomposiion as = + +, (12) where ( ) = + and L he lag operaor. The convergence of he mulivariae polynomial is assured in he sudied cases by he sabiliy of he esimaed VAR models. Noice ha he elemens of he vecor of errors are conemporaneously 11

13 correlaed so ha a shock in one variable ( 0) implies an insananeous shock in he res of variables. To avoid his problem and obain isolaed shocks on each variable, an orhogonalizaion of i is needed in order o obain he impulse responses. This model is op s d denoed as he Srucural VAR(k) model and is (srucural) shocks = (,, ) conemporaneously uncorrelaed. Here we denoe he srucural shock of OP, is inerpreed as an AS shock and as AD shock. The idenificaion of he srucural VAR model (or he orhogonalizaion of he reduced form errors) can be achieved in various ways; here we follow he Blanchard and Quah (1989) mehod which consiss of seing a se of resricions on he marix of long run effecs as = (1) = (13) According o he heoreical model in Eq. (9), real OP are free from domesic demand and supply shocks, i.e., he conemporaneous effecs of AS and AD shocks on OP are zero. Therefore, we have 2 more resricions: zero shor-run resricions on OP, = = 0. Finally, we impose a long-run resricion (Blanchard and Quah, 1989), where AD shocks have no long-run effecs on he growh of oupu, = 0. For simpliciy he srucural shocks are normalized o have uni variance so ha, denoing he resriced marix, hen ( ) =. To accoun for asymmeric effecs, his sudy separaes OP innovaions ino posiive and negaive pars following Mork (1989) 12. Mork (1989) inroduces an asymmeric concep of OPs and proposes o separae posiive from negaive OP changes. He indicaed ha here is an asymmery effec of OPs on macroeconomic variables. He confirmed ha posiive OP shocks have a significanly negaive impac on GDP while negaive OP shocks have no significan effecs for oil-imporing counries. He defines posiive and negaive quarerly OP OP ( Δ and ΔOP ΔOP Δ OP respecively) changes in he following ways; 1 max 0, OP OP (14) 1 min 0, OP OP (15) 12 The approaches of Hamilon (1996) and Lee e al. (1995) are no used due o he curren size of he paper and also due o he facs ha he resuls used by Mork s approach are more explanaory. 12

14 3.3 Impulse Response Analysis: Linear Specificaion The accumulaed response of each variable o a shock of magniude equal o wo sandard deviaion confidence inerval (dashed line) is ploed in Figures 1 o 6. Generally speaking, OP shocks seem o have some long-run effecs on he oupu growh in he four sample economies, bu he magniudes are no comparable wih he AS shocks. The OP shocks have, as expeced, a posiive effec on he growh of oupu and inflaion in Indonesia and Malaysia, alhough he effec is insignifican as growh is raised by only 0.5% and 0.8% afer he shock for Indonesia and Malaysia respecively, 13 see figure 1a-1b 14. This posiive oupu response is consisen wih he convenional wisdom ha an increase in OPs leads o a rise in he revenue and income of oil-exporing counries (see Io (2008), Bjornland (2000) and Rauava (2004). However, he confidence inerval confirms ha afer he firs quarer OP shocks have a negligible effec on oupu in he long-run. This may be because Indonesia recenly experienced an increase in he producion dependency on oil and has a high share of oil in he consumpion bundle. In addiion, he Indonesian economy has experienced sagnan oil producion when compared o demand due o he exhausion of old oil fields and lack of invesmen in exploraion for oil reserves. Hence, an increasing OP may have a negligible posiive effec on is oupu. [FIGURE 1] Figure 1: Accumulaed Effecs of OP Shocks on GDP Growh 13 This resul conradics ha of Salim and Rafiq (2011) findings where OP shocks end o lower GDP in Indonesia and Malaysia. 14 According o Hsing (2012), OP shocks would no damage he oupu of Indonesia because his counry produces mos of is oil o mee he domesic consumpion. 13

15 As an oil-imporing counry, OP shocks have a negaive and permanen effec on he oupu of Pakisan, where growh decreased by 0.3% afer he shock. This may be caused by he recen move owards oil dependen producion and he fac ha Pakisan impors around 80% of is domesic demand. The confidence inervals (see figure 1c) verify ha he effec on oupu of OP shocks is no significan in Pakisan. [FIGURE 2] Figure 2: Accumulaed Effecs of OP Shocks on inflaion For oil-imporing counries like Thailand, OP shocks have a negaive bu insignifican effec on he growh of oupu, where he negaive response of oupu o OP shock is abou 0. 8% (figure 1d). The reason behind he small damage of OP on GDP may be because Thailand 14

16 produces a sizeable amoun of oil which mees some par of domesic demand even hough i is a ne oil-imporing counry. Furhermore, he Thai governmen no only subsidizes OP bu also promoes expors, which brings a massive amoun of rade surplus. Furhermore, he negligible response of oupu o OP shocks (in Thailand) is also affeced by he paern of inernaional rade. Abeysinghe (2001) argues ha a higher OP can affec open economies direcly and indirecly on which he indirec effec runs hrough rading parners of he economy. During he bad period, a par of rade surplus can be used o overcome he increased world OP. The crucial domesic demand could be auarkic. I should also be noed ha if daa from longer periods are available, he long-run oupu response may well be revealed as negaive in alignmen wih he expecaions for oil-imporing counries in heory. Oil price shocks have posiive effecs on domesic prices, which cause inflaion in all counries. As expeced, we find ha he oil-imporing counries (Pakisan and Thailand) experience a significan boos in inflaion afer an OP shock as repored in figures 2c-2d, because a higher OP leads o an increase in he cos of producion. In conras, oil-exporing counries are responding o he same shock posiively due o an increase in naional income hrough greaer oil-expor revenue, which implies ha high OP increases he growh of demand from oil producers. However, as indicaed in figure 2a-2b, he response of domesic price growh o OP shocks is negligible for Indonesia and Malaysia, which can be aribued o he response of exchange raes, which end o appreciae in oil-exporing counries, exering a non-significan effec on inflaion 15. [FIGURE 3] Figure 3: Accumulaed Effecs of AS Shocks on Oupu Growh 15 See Baumeiser e al. (2010). 15

17 As expeced, he reacion of oupu o AS shocks is posiive and permanen in all counries (see figure 3). The impac of a produciviy shock vary from 2.2% o 3.8%. However, he immediae reacions o AS shock flucuae from period o period for each economy. The posiive effecs become sable afer 4 quarers for all counries. On he oher hand, he AS shocks have a relaively sable negaive impac on price growh in all invesigaed counries (figure 4). In Indonesia, he cumulaive negaive response of domesic price growh reaches a rough afer 1 quarer (around 2%) and becomes sable afer 12 quarers a 4%. In Pakisan, he domesic price growh is reduced by 0.5%. The AS shocks have a significanly negaive effec on domesic prices and cause deflaion in Thailand. The price response o he AS shock are relaively smaller in Malaysia when compared o he oher hree counries. This is perhaps due o he fac ha produciviy growh in he radable secor causes high domesic AD (higher expor) as a resul of less deflaion. Demand (inflaionary) shocks have a significanly posiive effec on he oupu in all counries, bu he posiive effec gradually decreases and dies ou afer 10 quarers. In he long-run, he response of GDP o inflaionary shocks disappears in all counries, which is consisen wih he heoreical expecaions and resricions (e.g. Blanchard and Quah, 1989). [FIGURE 4] Figure 4: Accumulaed Effecs of AS Shocks on Inflaion 16

18 The inflaionary shocks have a permanen and posiive effec on inflaion in all counries, which is consisen wih he economic heory. The response is higher in Indonesia and Pakisan, where a single uni shock corresponds o abou a 3% and 2.2% rise in price growh respecively. In comparison, he demand shocks caused only a 0.6% increase in price growh in Malaysia. Thailand is in beween, wih around 1.1%. According o he findings, he impac of OP shocks on seleced Asian developing economics is no as high as in indusrialized counries highly dependen on oil for insance he US, Japan, Russia, UK, Norway and China ec., which are among he bigges oil users in he world (Abeysinghe, 2001; Hamilon, 1983; Bjornland, 2000). Compared o large and esablished economies, he counries seleced for his sudy suffer less afer OP shocks which may be due o heir low dependency on oil for producion and small conribuion of oil in he consumpion bundles. In addiion, i can be aribued o direc governmen subsidies o he OP o proec ciizens and domesic indusry from inernaional OP shocks. The subsidized OP benefis many economic secors and herefore effecively conrols adverse effec of OP in he sample counries. The evidence suggess ha he relaionship beween OP and GDP is negligible. [FIGURE 5-6] Figure 5: Accumulaed Effecs of AD Shocks on Oupu Growh 17

19 Figure 6: Accumulaed Effecs of AD Shocks on Inflaion 3.4. Impulse Response Analysis: Asymmeric Specificaion To assess asymmeries wih respec o he sign of he OP shock under differen saring condiions, cumulaive responses of oupu growh o a posiive and o a negaive OP shocks are obained by using a VAR. Figure 7a-7h shows he cumulaive response of oupu growh o OP shocks, where he lef panel shows he responses o a posiive OP shock and he righ panel presens he responses for a negaive shock. For oil-exporing counries, posiive OP shocks induce posiive effecs on oupu growh and inflaion in Indonesia and Malaysia however he impac is insignifican in Malaysia. In conras o he linear model, he response of oupu growh o an OP shock is highly significan for Indonesia. Oupu growh falls in response o a negaive OP shock in oil-exporing counries and our findings confirm ha he magniude of he response is significanly larger han wih he linear model esimaion. The 18

20 fall in oupu growh resuling from a negaive OP shock is greaer han he increase in oupu resuling from a posiive OP shock. Overall, as expeced, he responses of growh and inflaion are asymmeric wih respec o he sign of he OP shocks for boh oil-exporing counries. Addiionally, he asymmeric effecs of OP on oupu growh are larger han he symmeric effecs. Therefore, hese findings are in line wih hose in previous sudies (An e al., 2014; Kilian and Vigfusson, 2011; Mendoza and Vera, 2010; Moshiri, 2015). For boh counries (Indonesia and Malaysia) a non-linear specificaion yields a more accurae represenaion han he linear model. Wih regards o he oil-imporing counries, in Pakisan he reducion in oupu growh resuling from a posiive OP shock is larger han he increase in oupu growh due a negaive OP shock. However, for Thailand he response of oupu growh o a negaive OP shock is larger han he absolue reacion o a posiive OP shock and in general, he response of growh o OP shocks is significan in Thailand. The findings are consisen wih hose in he previous lieraure on oil-imporing counries. (Jiménez-Rodríguez and Sánchez, 2005; Hamilon, 2009, 2011; An e al., 2014). The findings for Thailand are in line wih he resuls of Köse and Baimaganbeov (2015), who find a greaer impac of negaive OP shocks on oupu growh. The evidence shows ha he non-linear approach produces impulse response funcions ha are comparably more precise han hose generaed hrough he linear model. Figure 7: Accumulaed Effecs of Posiive and Negaive OP Shocks on GDP Growh: Asymmeric Specificaion Posiive OP Changes Negaive OP Changes 7a. 7b Indonesia Indonesia c. 7d. 19

21 Malaysia Malaysia e. 7f Pakisan Pakisan g. 7h Thailand Thailand Figure 8a-8h represens he impulse response funcions of inflaion o an OP shock. In he case of posiive OP changes, we observe ha despie each counry responding somewha di erenly o OP shocks, here is a similar paern in he impulse response funcions. For posiive OP innovaions, he response of inflaion is consisen wih he sign of he OP shock in all sample counries, bu only significan for Indonesia. Furhermore, in oil exporing counries, he reducion in inflaion due o a negaive oil price shock is greaer in magniude han he increase in inflaion resuling from a posiive oil price innovaion. In addiion, he increasing response of inflaion due o a posiive OP shock in he non-linear specificaion is 20

22 consisen wih he linear specificaion in all counries however, he magniude of he shock is greaer in he former case. These findings are consisen wih hose of Hooker (2002). Surprisingly, we have found ha negaive OP shocks cause inflaion in he oil-imporing counries of Pakisan and Thailand. This resul can be raionalised in erms of he unusual circumsances experienced by hese economies during he period under sudy. In conras o he linear specificaion, inflaion responses o negaive OP shocks are significan in all of he counries, excep Malaysia. Figure 8: Accumulaed Effecs of Posiive and Negaive OP Shocks on Inflaion: Asymmeric Specificaion Posiive OP Changes Negaive OP Changes 8a. 8b Indonesia Indonesia c. 8d..020 Malaysia.012 Malaysia e. 8f. 21

23 Pakisan Pakisan g. h..04 Thailand.020 Thailand Forecas Error Variance Decomposiion: Linear Specificaion In his subsecion, we sudy he forecas error variance decomposiions (FEVDs), which allow us o verify how much of he forecas error variance is explained by shocks o each explanaory variable in a sysem over a ime period. Variance decomposiion is based on srucural decomposiion (orhogonalizaion) esimaed in he facorizaion marices for he idenified VAR model. For each counry in his sudy, variance decomposiion is used o measure he proporion of flucuaions in oupu caused by OP, AS and AD shocks. Tables 2-5 show he variance decomposiions of he linear specificaion. [TABLE 2] Table 2: Forecas-Error Variance Decomposiions of Oupu Growh and Inflaion Indonesia: Linear Specificaion Horizon Oupu Growh Inflaion (quarers) OP shock AS shock AD shock OP shock AS shock AD shock

24 The FEVDs for Indonesia are repored in Table 2. These resuls show ha real OP shocks cause 2% of he variaion in oupu a all horizons. AS shocks conribue o 41% of he changes in oupu in Indonesia in he shor-run, whils his decreases o abou 33% in he long-run. The ineresing finding is ha nominal AD (inflaionary) shocks is he main source of variaion in oupu and explains abou 64% of flucuaions in oupu growh. The finding is very much consisen wih he hyperinflaion period during he Asian financial crisis, where he rae of inflaion rose sharply and reached abou 80% in mid In response, Indonesian GDP growh quickly decreased winessing negaive economic growh of over 13% in Oil price disurbances explain close o 0% of he variaions of inflaion in he long-erm, while AS disurbances explain abou 49% o 69% of forecas-error variance of inflaion in Indonesia. In he shor-erm, inflaionary shocks conribue abou 31% o he variaion in prices bu his percenage gradually increases wih ime, reaching abou 51%. Table 2 clearly shows ha in he 12 h quarer hese AD shocks only explain 51% of long-run variaion in prices. Unlike he findings of Cover e al. (2006), who find ha he major source of long-erm inflaion variaion is AD shocks, he curren resuls indicae ha in Indonesia AS shocks cause he greaes changes in inflaion. In Malaysia, OP shocks accoun for 1% o 15% of he forecas-error variance in oupu as indicaed in Table 3, which also becomes sable in a longer-erm. The conribuion of supply shocks o oupu varies wihin he range of 81%-88%. Regarding long-erm inflaionary change in Malaysia, abou 3% of his is explained by real OP shocks. Demand disurbances are he single mos significan source of variaion in inflaion, accouning for abou 88% of he variance. Supply shocks only explain 9% of hese flucuaions in he long run. [TABLE 3] Table 3: Forecas-Error Variance Decomposiions of Oupu Growh and Inflaion Malaysia: Linear Specificaion Horizon Oupu Growh Inflaion (quarers) OP shock AS shock AD shock OP shock AS shock AD shock

25 In Pakisan, OP shocks have a relaively small effec on oupu as exhibied in Table 4, and conribue o only 6% of he long-run variaion in oupu. AS shocks are he imporan causal facors of oupu variaion conribuing o 90% of he variance over he ime period under invesigaion. Nominal AD shocks explain only 4% of oupu flucuaions in he longrun. The resuls show ha he real OP shock causes 1% of shor-erm and abou 15% of longerm variaions in price growh. In he shor-erm, demand disurbances explain abou 85% of he variaions in domesic prices, wih his conribuion gradually declining as ime goes on. Supply disurbances are less imporan for explaining inflaion changes in Pakisan, conribuing only 5% in he long-run. [TABLE 4] Table 4: Forecas-Error Variance Decomposiions of Oupu Growh and Inflaion Pakisan: Linear Specificaion Horizon Oupu Inflaion (quarers) OP shock AS shock AD shock OP shock AS shock AD shock The forecas-error variance decomposiion for oupu and price growh in Thailand is repored in Table 5. [TABLE 5] Table 5: Forecas-Error Variance Decomposiions of Oupu Growh and Inflaion Thailand: Linear Specificaion Horizon Oupu Inflaion (quarers) OP shock AS shock AD shock OP shock AS shock AD shock

26 In erms of variance decomposiion of oupu, AS shocks conribue he bigges proporion in Thailand. The shor run effec is 97% which gradually decreases wih he forecas horizon o 92%. In he long-run, OP and AD shocks accoun for 4% and 5% of he variaion of oupu respecively. Supply shocks in Thailand accoun for a sizable flucuaion in inflaion represening abou 16% of he long-erm variabiliy. Real OP disurbances explain abou 16% of he variaion in inflaion and demand disurbances describe abou 68%. To summarise he FEVD resuls, he sudy concludes ha AS disurbances are he mos imporan facor behind oupu movemens in boh shor-erm and long-erm in his se of counries excep in Indonesia. I seems ha he inflaionary (AD) shocks are he main source variaion in oupu for Indonesia. I suggess ha he recession in he Indonesia during was dominaed by inflaionary disurbances. For all four counries, oupu behaviour seems consisen wih real business cycle heory. Demand is he main deerminan of he variabiliy in domesic prices in Malaysia, Pakisan and Thailand. In Indonesia, supply shocks are more dominan han demand shocks in explaining domesic price flucuaions. In all counries, he demand, and supply shocks are a significan source of business cycles Forecas Error Variance Decomposiion: Asymmeric Specificaion Tables 6 and 7 show he findings of he FEVD wih regards o he non-linear model, which presens how much of he unanicipaed changes of he oupu and inflaion are explained by OP shock. The conribuion of posiive OP shocks o oupu and inflaion variabiliy is 7% and 8% respecively for Indonesia, as repored in Table 6, whils negaive OP innovaions explain 9% of he variaion in boh oupu and inflaion. In he case of Malaysia, negaive OP shocks conribue sizably o variaions in oupu (23%) and inflaion (20%). We nd ha shocks o OP variables end o explain more of oupu and inflaion volailiy in asymmeric han in symmeric speci caions in boh oil-exporing counries. TABLE 6 Table 6: Forecas-Error Variance Decomposiions of Oupu Growh and Inflaion Indonesia and Malaysia: Asymmeric Specificaion Indonesia Malaysia Horizon Oupu Growh Inflaion Oupu growh Inflaion Quarers OP+ OP- OP+ OP- OP+ OP- OP+ OP- 25

27 For he oil -imporing counries Pakisan and Thailand, posiive OP shocks explain subsanial variaions in inflaion, 16% and 19% respecively, as presened in Table 7. In he case of Thailand, he conribuion of negaive OP shocks o variaions in oupu is 19%. Overall, he findings confirm ha for models using a non-linear specificaion of he oil price, OP shocks end o explain more of oupu and inflaion variabiliy han models using a linear specificaion in all sample counries under sudy. TABLE 7 Table 7: Forecas-Error Variance Decomposiions of Oupu Growh and Inflaion Pakisan and Thailand: Asymmeric Specificaion Pakisan Thailand Horizon Oupu Growh inflaion Oupu growh inflaion Quarers OP+ OP- OP+ OP- OP+ OP- OP+ OP CONCLUSIONS This paper examines he dynamic effec of OP, AD and AS shocks on boh oupu and domesic aggregae price growh in Indonesia, Malaysia, Pakisan and Thailand using a srucural VAR model wih a mixure of shor-run and long-run resricions following he sandard Blanchard-Quah approach. Furhermore, we have compared he resuls using wo differen specificaions of he OP, namely a sandard linear model and a non-linear specificaion as proposed in he lieraure. The cumulaive impulse responses and forecas error variance decomposiion funcions are compued for each counry. Our resuls show ha in he linear model he effec of OP shocks on oupu is posiive in Indonesia and Malaysia, bu negaive in Pakisan and Thailand. However, he effecs are 26

28 insignifican for all counries. These resuls are consisen wih he findings of Ran and Voon (2012), who concluded ha OPs do no have a significan effec on economic aciviy in Eas Asian counries. When using he nonlinear model, in conras o he linear model, we find ha posiive and negaive OP shocks separaely have significan effecs on oupu and inflaion in Indonesia. Addiionally, we observe ha OP shocks have a significan impac on inflaion in Pakisan and Thailand. Overall, he findings confirm ha OP changes have larger effecs on macroeconomic variables when a non-linear model is used as opposed o he sandard linear model in all four economies. The evidence also shows ha posiive OP innovaions have di eren impacs on oupu growh and inflaion o hose of negaive innovaions. This conrass wih he linear model in which OPs are assumed o have symmerical impacs on real aciviy. The findings of he curren sudy also sugges ha OP increases causes inflaion in all counries, which corresponds o an earlier conjecure in secion 2. This is consisen wih he finding of Abeysinghe (2001) and Cunado and Perez de Gracia (2005) ha even oil-exporing counries such as Indonesia and Malaysia couldn evade he adverse repercussions of high OPs. The inflaionary consequences for oil-exporers are negligible, possibly because of he exchange rae appreciaion in he afermah of an OP shock. The resuls of OP shocks are paricularly negaive for oil-imporers, as hey may bear losses because of he high inpu cos. On he oher hand, oil-exporers may experience some benefis and revenue increases bu also suffer high inflaion. Alhough our variance decomposiion invesigaion suggess ha OPs have a role in explaining oupu and inflaion volailiy in all sample economies, our findings sugges ha OP shocks are more likely o cause flucuaions in he macroeconomic variables of some counries han in ohers. The relaively small or negligible effec found for some counries could be aribued o governmen direc conrol and subsidised OPs which help o minimize he adverse effec of OP on real aciviies and o avoid a sharp decline in GDP and high inflaion in response o posiive OP shocks. In addiion, he economic srucure of a counry plays a crucial role, which may affec he influence of OP shocks on macroeconomic flucuaions. Counries wih a low producion reliance on oil, a low share of oil in he consumpion bundle and relaively low labour inensiies in producion are expeced o suffer less from OP shocks. Shocks of AD demand and AS have played a significan role in explaining recessions, paricularly he recession in Eas Asian counries. For he recession experienced in he OP had lile impac and he recession was largely caused by oher disurbances in all counries. 27

29 In alignmen wih convenional wisdom, produciviy (AS) shocks are he key reason for flucuaions in oupu and nominal AD shocks are he main facor inducing flucuaions in aggregae price growh in all counries excep in Indonesia. In Indonesia, demand disurbances are more imporan han supply disurbances in explaining oupu flucuaions and ineresingly, produciviy shocks are he main reason for flucuaions in aggregae price growh in Indonesia because is economy is highly dependen on expors. These findings may sugges ha oupu growh of he rade secor is higher han ha in non-rade secors in Indonesia. The findings confirmed all he resricions imposed by he model o idenify he differen shocks. Mos of he dynamic changes of he macroeconomic variables are in line wih he economic model, and he shocks fi well wih acual evens ha have occurred in he various counries. These findings carry significan policy recommendaions o governmens, suggesing he need for comprehensive reforms of OP and oil subsidy policies. The governmen policies should pay more aenion o emerging new and advanced soluions o achieve OP sabiliy. Many counries increase oil subsidies in order o minimize he damaging effecs of OP increases on poor people however he consequences of OP sabilizaion policies are budge defici 16 and polluion. The policy should be implemened in such a way so as o reduce OP shocks o he economy and minimize budge defici. This suggess ha he removal of oil subsidies should be gradual because people could recognize policies and become more resilien o OP shocks. Furhermore, in order o conrol inflaion, he moneary policy response o an OP shock could involve raising he ineres rae o reduce consumer and invesmen spending. Furhermore, such resuls have oher significan implicaions, in paricular ha he idenificaion and decomposiion of demand and supply shocks are crucial. This enables beer analysis of he effecs of moneary and fiscal policy on he economy, which is in urn aken as a measure of progress of he growh and improvemen of marke mechanisms. I could also help hese counries walk ou of budge defici and achieve a balanced budge, which will subsanially reduce he risk of falling ino financial crisis again. Overall, for a fuller undersanding of recen developmens in he economies ha have been considered, and in order o srucure economic policies ha aim a sabilizing he macroeconomy, he resuls of he curren sudy should be carefully aken ino accoun during 16 Rafiq e al. (2009) examined he impac of OP shocks on economic aciviy in Thailand by employing a VAR model. Using quarerly daa from 1993q1 o 2006q4 hey found a srucural break in he ime series daa during he Asian financial crisis of They also found ha he budge defici originaed mainly from movemens in OPs during he pos crisis period, which may have been due o he floaing exchange rae policy. 28