Workshop on “Money, Finance and Banking in East Asia” Training Centre of the Deutsche Bundesbank, Eltville 5-6 December 2011 Mardi Dungey University of Tasmania and CAMA Presentation to “Modelling East Asian economies in a small open economy VECM: the influence of international and domestic shocks“ www.bundesbank.de
Modelling East Asian economies in a small open economy VECM: the in‡uences of international and domestic shocks Mardi Dungey � % and Tugrul Vehbi % � University of Tasmania and CAMA,ANU % CFAP, Cambridge December 2011 1/31
Motivation � Apply recent macroeconometric modelling techniques to ASEAN economies � address the issue of the source of international shocks in the region � We want to relate to contemporary modelling techniques � DSGE models particularly for New Keynesian theory � structural VAR for empirical dynamics � There is an identi…cation problem in open economy modelling in DSGE and VAR models � often resolved by a small open economy assumption as here � longer term agenda is to move to interdependent economies 2/31
Aspects of the Solution � Use contemporary NK theory as basis of restrictions � A SVAR framework to capture dynamics � explicit modelling of the long run � separation of long and short run shocks � Apply this framework to 5 ASEAN economies � Singapore, the Philippines, Thailand, Malaysia, Indonesia � Foreign e¤ects represented by US or China 3/31
Outline 1 Introduction and some literature 2 A basic theoretical framework 3 Econometric speci…cation 1 Permanent versus transitory shocks 2 VECM speci…cation 4 Data 5 Results 1 impulse response functions 2 historical decompositions 6 Concluding remarks 4/31
Existing empirical literature � 100s of VAR studies on the US, closed economy � classic benchmarks are Sims (1980, 1992) � Common …ndings are: � price puzzle: tighter monetary policy does not result in lower in‡ation � exchange rate puzzle: increaeses in domestic interest rates do not result in appreciation of US dollar � but these are worked around and seem to work in general quite well � New Keynesian DSGE models � largely Bayesian estimations: eg � Christiano, Eichenbaum, Evans (2005), Lubik and Schorfheide (2005), del Negro and Shorfheide (2008) � Calvo pricing, staggered contracts � Gali and Monacelli (2005) and Monacelli (2005) 5/31
Existing empirical literature ASEAN economies � Chow and Yoonbai (2003) 3 variable VARs in output � Zhang et al (2004) 3 variable VAR for demand, supply, monetary policy shocks � Huang and Feng (2006) 4 variable VAR, …nd some commonality amongst countries � Zhang et al (2010) closest to us � structural VAR with exogenous US shocks � …nd US shocks to be a dominant in‡uence 6/31
A Basic Theoretical Framework � A stylized small open economy model IS curve, NK Phillips curve, monetary policy reaction function, UIP condition y t = µ E t ( y t + 1 ) + ( 1 � µ ) y t � 1 + φ ( r t � E t � 1 π t ) + θ 1 ∆ q t + θ 2 y � t + ǫ AD , t π t = δ E t π t + 1 +( 1 � δ ) π t � 1 + λ y t + θ 3 ∆ q t + ǫ AS , t r t = ρ r t � 1 + ( 1 � ρ )( β E t π t + 1 + γ y t ) + ǫ MP , t ( r t � E t π t + 1 ) � ( r � t � E t π � E t ( ∆ q t + 1 ) = t + 1 ) � ǫ RER , t y t ( y � t ) domestic (foreign) output gap : r t domestic nominal interest rate : π t domestic in‡ation rate : q t real exchange rate : 7/31
Empirical framework � We want to use a SVAR approach building on the theoretical relationships � Want SVAR for empirical coherence, allows better dynamics � Innovation: � using the properties of the data (empirical and theoretical) to provide identi…cation � accounting for changes in exchange rate regime in some ASEAN economies in 1997/1998 8/31
Empirical Framework � Properties of the data � We know that y t , q t , y � t will be I(1) � In fact from the IS equation we know they should cointegrate � Therefore we will have a mix of permanent and transitory shocks in the system � This leads us naturally to a SVECM framework � but we will need to be able to encompass I(0) and I(1) variables within it � solution suggested by Pagan and Pesaran (2009) 9/31
Empirical Framework � SVECM B ( L ) ∆ Y t = Π Y t � 1 + ε t = αβ 0 Y t � 1 + ε t with B 0 nonsingular and E ( ε t ε 0 t ) diagonal 2 t ) 0 has r cointegrating � Partition the n � 1 vector Y t = ( Y 0 1 t , Y 0 vectors � Y 1 t is (( n � r ) � 1 ) which experience permanent shocks � Y 2 t is ( r � 1 ) which experience temporary shocks 10/31
Empirical Framework � Common trends representation ∆ Y t = F ( L )( B 0 ) � 1 ε t where F ( L ) = I n + k + F 1 L + F 2 L 2 + ... � and F ( 1 ) = F � � α 0 α � 1 F = β ? ? Ψ ( L ) β ? ? , ? α = 0 , β 0 ? β = 0 , F α = 0 and β 0 F = 0 . with α 0 � Practically what does this mean? � means we can partition the matrix and …gure out what happens in the case of permanent and transitory shocks 11/31
Empirical Framework � Take the …rst ( n � r ) permanent shocks represented with ε 1 jt and the ε 2 jt to be transitory � ε 1 t � ∆ Y t = F ( L )( B 0 ) � 1 ε 2 t 12/31
Empirical Framework � Take the …rst ( n � r ) permanent shocks represented with ε 1 jt and the ε 2 jt to be transitory � ε 1 t � ∆ Y t = F ( L )( B 0 ) � 1 ε 2 t � So we know the e¤ects of the transitory shocks on ∆ Y t = 0 so � 0 ( n � r ) � r � F ( B � 0 ) � 1 = 0 I r + k 12/31
Empirical Framework � Take the …rst ( n � r ) permanent shocks represented with ε 1 jt and the ε 2 jt to be transitory � ε 1 t � ∆ Y t = F ( L )( B 0 ) � 1 ε 2 t � So we know the e¤ects of the transitory shocks on ∆ Y t = 0 so � 0 ( n � r ) � r � F ( B � 0 ) � 1 = 0 I r + k � So we can write this to show that � 0 ( n � r ) � r � � α � � 1 R = B � 0 α R = α � R = α � 2 R I r + k The only way to satisfy this is if α � 1 = 0 12/31
Empirical Framework � Take the …rst ( n � r ) permanent shocks represented with ε 1 jt and the ε 2 jt to be transitory � ε 1 t � ∆ Y t = F ( L )( B 0 ) � 1 ε 2 t � So we know the e¤ects of the transitory shocks on ∆ Y t = 0 so � 0 ( n � r ) � r � F ( B � 0 ) � 1 = 0 I r + k � So we can write this to show that � 0 ( n � r ) � r � � α � � 1 R = B � 0 α R = α � R = α � 2 R I r + k The only way to satisfy this is if α � 1 = 0 � MEANS: that transitory shocks may have a non-zero error correction term, permanent shocks must have a zero error correction term 12/31
Empirical Framework � One further important aspect: 13/31
Empirical Framework � One further important aspect: � exchange rate regime changes are handled with an interactive dummy variable speci…cation for the break B ( L ) ∆ Y t + B � ( L ) D t ∆ Y t = Π 1 Y t � 1 + D t Π 2 Y t � 1 + ε t where � 0 : before regime change D t = 1 : after regime change 13/31
Application: the long run � from the IS equation we should have cointegrating vector between � y t ASEAN country GDP � y � t US GDP � q t real exchange rate � This is 3 � I(1) variables, with 1 cointegrating vector ) 2 permanent shocks � Assume these originate in y t and y � t 14/31
The structural form: � f y � t , π � t , r � t , y t , π t , r t , q t g , augment the Phillips curve with exogeneous oil price in‡ation 2 3 1 6 7 b 0 1 6 7 21 6 7 ∆ Y t = αβ 0 Y t � 1 b 0 1 6 7 32 4 5 b 0 b 0 1 42 43 b 0 b 0 b 0 b 0 1 51 52 53 54 2 3 2 3 b l 0 0 0 0 0 11 6 7 6 7 b l b l b l b l b l c 6 7 6 7 21 22 23 24 25 6 7 6 7 b l b l b l + 0 0 ∆ Y t � 1 + 0 oil t + ǫ t 6 7 6 7 32 33 35 4 5 4 5 b l b l b l 0 0 0 42 43 44 b l b l b l b l b l 0 51 52 53 54 55 15/31
Application: the long run � where 2 3 0 0 0 2 3 6 7 0 0 0 β 11 1 0 0 β 51 6 7 αβ 0 = 6 7 4 5 0 0 0 0 1 0 0 α 32 6 7 4 5 0 α 42 α 43 0 0 1 1 0 α 51 α 52 α 53 � note that UIP is not imposed as there is little empirical support in the literature 16/31
Data � Variables list: f y � t , π � t , r � t , y t , π t , r t , q t g � Sample period: 1986Q1 to 2009Q4 � Estimated with 3 lags in levels (2 lags in changes) � dummy variable added for 1997Q3 to 1998Q4 � Show the example of Singapore for the impulses � Historical decompositions for Singapore, Thailand, Malaysia � In‡uence of foreign shocks from US and China for all countries 17/31
Singapore 18/31
Impulse responses for foreign output shock in Singapore 19/31
Historical Decomposition: Singapore output 20/31
Historical Decomposition: Singapore in‡ation 21/31
Recommend
More recommend
