1Agata STĘPIEŃ Bilge Kagan OZDEMIR Renata SADOWSKA Winfield TURPIN Structural VAR Modelling Of Monetary Policy For Small Open Economies: The Turkish CaseAgata STĘPIEŃBilge Kagan OZDEMIRRenata SADOWSKAWinfield TURPIN
3- on account of the effects of monetary policy shocks - VAR methodology:Produces efficient results for small closed economiesProvides uncertain empirical results for small open economies- on account of the effects of monetary policy shocks -
4AIM: to present why SVAR methodology is better than VAR We investigate the utility of the structural VAR approach in conventional empirical puzzles:The price puzzleThe liquidity puzzleThe exchange rate puzzle
5Non-recursive VAR’s are called structural VAR (SVAR) models. Empirical PuzzlesResult from the recursive structure implied by the standard identification procedure of VAR modelsNon-recursive identification schemes effectively solve these puzzles:Non-recursive VAR’s are called structural VAR (SVAR) models.
6Price Puzzle Sims (1992)In various empirical VAR studies, a contractionary monetary shock causes a persistent increase in price level rather than a decrease.This odd response of the price level to a restrictive monetary policy shock is called “the price puzzle”
7The Liquidity Puzzle Leeper & Gordon (1992) A similar anomaly has been observed in the response of interest rates to a shock to monetary aggregates. Following an expansionary shock to the money variable, the interest rate exhibits a positive response creating “the liquidity puzzle”.
8The Exchange Rate Puzzle Grilli and Roubini (1995) & Sims (1992) In an open economy environment a positive innovation in interest rates seems to result in a depreciation of the local currency rather than an appreciation. This is “the exchange rate puzzle”.
9The dataAll of our estimations use monthly data for Turkey covering the period 1997:1 to 2004:12IPI : Industrial production indexP : Wholesale price indexM : Monetary aggregate (M1)R : Short-term interest rates (overnight rates)REDEX : Real effective exchange rate indexEX : Nominal exchange rateAll variables are in logarithm levels except the short-term rate.
11Structural VAR methodology pth order reduced form VAR:yt - n x 1 vector of endogenous variablesAi - the coefficient vector of lagged variables yt - pet - the vector of serially uncorrelated reduced form errorswith (etet`) = Σthe more compact form:A(L) - a matrix polynomial in the lag operator L
12the structural form of VAR: where:B(L) - a pth order matrix polynomial in the lag operatorut - nx1 vector of structural innovations, with:ut – serially uncorrelated and diagonalThe relationship between the structural and the reduced modelB0A(L)=B(L)B0e=uΣ=(B0-1)Ω(B0-1)
13Imposing parameter restictions Cholesky decomposition - orthogonalizing the covariance matrix of reduced form residuals gives an exactly identified system,implies a recursive structure among the variables of the system.structural VAR- allows us to use a non-recursive structure- we identify the model by imposing short-run restrictions on B0, or long-run restrictions on B1Kim and Roubini (2000): indentification = at least n(n+1)/ restrictions on B0
14Determining the set of restrictions on B0 2 approaches:(i) an explicit macroeconomic model (Gal (1992))(ii) choosing restrictions based on the structure of the economy ((Leeper et al. (1996) and Kim and Roubini (2000)).- restrictions, which produce the results consistent with economic theories,- restrictions, which are not rejected by data.
16Lag order selection FPE - the final prediction error, . varsoc R lIPI lOP lM lP lREDEXSelection order criteriaSample: 1997m m Number of obs =|lag | LL LR df p FPE AIC HQIC SBIC || || 0 | e || 1 | e * * || 2 | e || 3 | e-13* || 4 | * e * |Endogenous: R lIPI lOP lM lP lREDEXExogenous: _consFPE - the final prediction error,AIC - Akaike's information criterion,BIC - the Bayesian information criterion,HQIC - the Hannan and Quinn information criterion
17VAR model - results . var R lIPI lOP lM lP lREDEX, lag(1/3) Vector autoregressionSample: 1997m m No. of obs =Log likelihood = AIC =FPE = 3.24e HQIC =Det(Sigma_ml) = 2.70e SBIC =Equation Parms RMSE R-sq chi2 P>chi2RlIPIlOPlMlPlREDEX
24The stability of the model . varstableEigenvalue stability condition| Eigenvalue | Modulus || || i | || i | || i | || i | || i | || i | || i | || i | || | || | || i | || i | || i | || i | || i | || i | || i | || i | |All the eigenvalues lie inside the unit circleVAR satisfies stability conditionThe stability of the model
25Lagrange Multiplier test for autocorrelation in the residuals of VAR model . varlmarLagrange-multiplier test| lag | chi2 df Prob > chi2 || || 1 | || 2 | |H0: no autocorrelation at lag order
48Why SVAR is better than VAR? VAR MODELS:it is often difficult to draw any conclusion from the large number of coefficient estimates in a VAR system,vector autoregressions have the status of „reduced form'' and, thus, are merely vehicles to summarize the dynamic properties of the data,the parameters do not have an economic meaning and are subject to the so-called „Lucas critique'‘.SVAR MODELS:SVAR’s do not contain fixed-coefficient expectational rules. They are best thought of as giving linear approximations to the behavior of the private sector and monetary authorities. The private behavior they model thus implicitly includes dynamics arising from revision in forecasting rules as well as other sources of dynamics.