The Academy of Economic Studies Bucharest DOCTORAL SCHOOL

The Academy of Economic Studies, Bucharest DOCTORAL SCHOOL OF FINANCE AND BANKING International Transmission of Monetary Policy Shocks The Case of Romanian Economy MSc Student: ELENA BOJEŞTEANU Supervisor: Professor MOISĂ ALTĂR Bucharest, 2006

Contents § Motivation § The analysis of comovements using PCA § Identifying monetary policy shocks using a reaction function for the ECB § The transmission of shocks – a SVAR approach § Concluding remarks and further research

Key References § Two-country models: - Mundell Fleming (1962); - Svensson and van Wijnbergen (1989); - Obstfeld and Rogoff (1995); - Chari, Kehoe and Mc. Grattan (1997, 2000); - Engel and Devereux (2003). § Empirical Investigations: - Cushman and Zha (1997); - Christiano, Eichenbaum and Evans (1996, 1998); - Betts and Devereux (1999); - Bordo and Murshid (2002); - Pesaran et al. (2005).

Romanian Economy § The last five years: improved performance in terms of economic expansion, strengthening disinflation, reduction in budget deficit and unemployment. § A small open economy: a degree of financial and commercial openness exceeding 70% in the last 8 years. Main trading partner: European Union (over 70% of the total for exports, and more than 60% of the total imports. Trading currency: more than 60% settlement of exports and imports in euro, and approx. 30%in US dollar. § Capital account liberalization schedule. Romania is increasingly integrating into world markets and more precisely, into European structures. § Question: how do the commercial and financial linkages affect the Romanian main economic indicators? Is there a comovement in the economic variables, a synchronization of business cycles, or Romanian is still in an incipient phase of integration?

Tracing comovements using Principal Component Analysis § The method generates a new set of variables, called principal components § PC are orthogonal to each other § Each principal component (pc) is a linear combination of the original variables § The coefficients of each of the linear combinations are called loadings (or weights) § The first principal component explains the greatest amount of the total original variance § The sum of the pc variances equals the total variance of the initial system

Tracing comovements using Principal Component Analysis GDP growth rates § § The first component loading for Romania is small, which shows a lower correlation with the rest of the system. Sayek and Selover (2002) state that the first principal component might be thought of as the business cycle followed by the Western nations. For the second component, Romania has a much higher loading than the rest of the countries. The low and positive Romania’s loading on the first component may be a sign that the macroeconomic evolution in this country was in general out of step with the rest of the group.

Tracing comovements using Principal Component Analysis Real interest rates § § A higher degree of comovement in the financial sector that in the real one. The first principal component explains a substantial portion of the behaviour of interest rates across countries and can be interpreted as the common element of real interest rates. An atypical pattern for Romania is not out of the question, considering the high correlation with the second component.

Possible Interpretations for Monetary Policy Shocks Three general strategies for isolating monetary policy shocks: § The recursiveness assumption – based on the estimation of a reaction function for the monetary authorities - Christiano (1996); - Christiano, Eichenbaum and Evans (CEE, 1996, 1997); - Clarida, Gali and Gertler (1997); - Cushman and Zha (1997). § The narative approach - Romer and Romer (1989) § Long-run neutrality of money - Pagan and Robertson (1995)

Identifying the monetary policy shocks using a reaction function for the ECB § An exogenous monetary policy shock, εt - formalized by CEE (1998) as being the disturbance term in an equation of the form: St = f(—t) + εt , where St is the instrument used by the monetary authority and f( —t) is a linear function that captures the policy makers’ responses to variations in different economic variables, as they are known at time t. § An augmented reaction function for the Euro area: it = (1 -ρ)·α+(1 -ρ)·β·E[πt+n] + (1 -ρ) ·γ·yt + ρ· it + εt .

Identifying the monetary policy shocks using a reaction function for the ECB Studies estimating the reaction function using data before EMU: Data: § ex-post available data § survey data Gerdesmeier and Roffia (2003); Gerlach(2003); Surico (2003); Carstensen and Colavecchio (2005). Methodology: § GMM for ex-post available data (a popular technique in the rational-expectation context (Clarida, 1998)). Problem: selection of instruments. § OLS for survey data. Problem: constructing the data.

Identifying the monetary policy shocks using a reaction function for the ECB Ex-post available data 1996: 01 -2005: 04 from ECB and Eurostat databases § Interest rates: the interbank ON interest rate, the 3 M EURIBOR, and the 10 Y government bond yield § Price indices: annualized HICP and alternatively the core inflation (HICP - All items, excluding energy, food, alcohol and tobacco) § Output gap: from three measures for potential GDP: a Hodrick-Prescott filter (the smoothing parameter equal to 1600 for quarterly data), a linear and a quadratic trend. The three methods yield fairly similar results. § Monetary aggregates: M 3; a money gap was also used (the deviation of money growth from the reference value of a constant growth of 4. 5% per annum) § Exchange rates: nominal and real effective exchange rate.

Identifying the monetary policy shocks using a reaction function for the ECB Ex-post available data § § GMM. The instrument set includes lagged values (up to 4 lags) of the interest rate, inflation and output gap. The results are not very sensitive in respect to the number of lags used as instruments, the J-statistic supports the over-identifying restrictions implied by the model. The standard errors were computed using the delta method. The J-statistic reported in the table is the minimized value of the objective function, p(J), the null hypothesis that the overidentifying restrictions are satisfied (Hansen’s J-test).

Identifying the monetary policy shocks using a reaction function for the ECB Ex-post available data

Identifying the monetary policy shocks using a reaction function for the ECB Survey data Sample period 1999: 1 -2005: 4 § Quarterly forecasts based only on real-time available information § Solid arguments in favor of using survey data: policy makers; with lags; - they are more suitable to capture the forward-attitude of the - variables (in particular the series for output) are only available - data are often subject to revisions and it may take some quarters before the final series are available. § Inflation: based on the data from the Survey of Professional Forecasters (SPF), measured by the latest available forecast for the current year § A measure of the state of real economy: the Economic Sentiment Indicator (ESI). This economic index appears to be more closely tied to the Governing Council’s interest rate decisions than other variables capturing real economic activity.

Identifying the monetary policy shocks using a reaction function for the ECB Survey data § ESI as a leading indicator for economic activity Studies estimating the reaction function using survey data: Carstensen and Colavecchio (2005); Gerdesmeier and Roffia (2005); Gerlach (2004); Sauer and Sturm (2003).

Identifying the monetary policy shocks using a reaction function for the ECB Survey data Sample period 1999: 1 -2005: 4 § § The first two specifications show the results without partial adjustment. Although this restriction is rejected by the data, the estimates correspond to the original Taylor coefficients. The constant term is found to be statistically insignificant, similar to the findings of Carstensen and Colavecchio (2005). The equation that best fits the data is considered to be [9], with an interest rate smoothing and no constant term.

Identifying the monetary policy shocks using a reaction function for the ECB Survey data § Identified monetary shocks

Identifying the monetary policy shocks using a reaction function for the ECB Survey data § § § The results show a greater weight attached to the output gap relative to inflation, a conclusion similar to that of the studies using ex-post data. For some specifications, the constant term is found to be statistically insignificant. The real time forward-looking specifications of the Taylor rule using the SPF forecasts denote a stabilizing behavior and provide a better description of the actual behavior of the central bank.

The transmission of shocks – a SVAR approach Description of the variables. Quarterly data comprising: § Inflation rate (pi_ro), calculated using log deviation of the CPI from the previous quarter; § Core inflation (core 1), CPI – all items excluding administrated prices; § Real interest rate (rr_ro), the difference between BUBOR 3 M and the inflation rate; § Real GDP growth rate (d(y_ro)); § Exchange rate appreciation (d(log(er))), as the log difference between the quarterly mean of the exchange rate and that of the previous period. The series are seasonally adjusted using TRAMO/SEATS (Demetra). The sample period, due to data availability for the European reaction function is 1999: 1 -2005: 4 (28 observations). CB and Eurostat databases.

The transmission of shocks – a SVAR approach Main three methods to identify the pure innovations: § The recursive approach (the triangular Choleski decomposition) § The structural approach as advocated by Sims and Bernanke § The long-term restriction approach (the Blanchard and Quah decomposition)

The transmission of shocks – a SVAR approach The models: § Model (A 1): rr_ro, er, mshock; § Model (A 2): core 1, er, mshock; § Model (A 3): pi_ro, er, mshock; § Model (B): y_ro, rr_ro, er, mshock. Isolating pure shocks by Choleski ordering: § Model (A 1): mshock er rr_ro; § Model (A 2): mshock er core 1; § Model (A 3): mshock er pi_ro. The Choleski ordering places the monetary shocks first, reaffirming their exogeneity towards the Romanian economic variables.

The transmission of shocks – a SVAR approach Isolating pure shocks: The Model (B) relies on an identification scheme which assumes that contemporaneously (within a quarter), the external monetary shock affects only the financial variables (the exchange rate and the real interest rate) and not the real activity in Romania.

The transmission of shocks – a SVAR approach Tests for selecting the number of lags § Model (A 1): rr_ro, er, mshock; § Model (A 2): core 1, er, mshock; § Model (A 3): pi_ro, er, mshock.

The transmission of shocks – a SVAR approach Tests for selecting the number of lags § Model (B): y_ro, rr_ro, er, mshock

The transmission of shocks – a SVAR approach

Concluding remarks § § The monetary shocks are isolated by estimating a reaction function for the euro area. A more appropriate method to identify the policy shocks is to use the information set available at the moment the decision is made, i. e. survey data. The empirical evidence does not support an impact of these shocks on the internal variables. - - The number of observations used for the estimation may be inappropriate for the analyses of monetary transmission, knowing that the monetary decisions affect the economy only with lags. For the major part of the analyzed period, Romania’s exchange rate regime was managed floating, but according to empirical findings and to IMF, it was a mixed regime in the form of sliding band. The theoretical results for the case of floating exchange rate may not hold if this assumption is not met. Moreover, the capital account has not been fully liberalized and the stages with the greatest impact on the balance of payments occurred in only in 2005. The financial openness is questionable before this period.

Concluding remarks § Not European business cycles and monetary innovations determine the internal economic indicators, but domestic economic and political developments. - § During transition period major internal disturbances affected the Romanian economy; these internal shocks include the effects of domestic political conflicts, economic and financial crises, domestic policy mistakes and so on. Apart from the obvious advantages incurred by the imminent accession, the unpredictable effect of the European monetary policy on the Romanian economic variables can trigger integration costs not dealt with so far. Further research: - Alternative methods for identification of monetary policy shocks (using the data-determined approach and the Blanchard and Quah decomposition). The reaction function for the ECB can be obtained by employing monthly data, using cubic splines on real GDP. In order to test the relevance of theme it is useful to analyze the transmission of the identified monetary policy shocks in other countries except Romania, namely the new EU member countries. It can also be tested whethere is an asymmetry bween the effects of a negative and positive monetary shock.

Selected references