Academy of Economic Studies Doctoral School of Finance and Banking New Member States of the European Union on the road to monetary integration. The case of Romania MSc Student: Dumitru Daniel Supervisor: Professor Mois Altr Contents Motivation and objectives Optimal currency area and clustering analysis literature review Models used and estimation methodology Data Empirical results Conclusions Motivation The accession of the NMS to the European Union has stimulated a growing academic and policy debate about when should the euro area be extended to the new EU members based on the achievements in the convergence process. The current international financial and economic crisis triggered by the subprime mortgage market from United States led to a rethinking in the euro adoption strategies for some New Member States (NMS) in order to speed-up the process. Nevertheless, the euro adoption is still constrained by the fulfillment of the convergence criteria: nominal convergence and real convergence. Nominal convergence (Maastricht convergence criteria) Real convergence A special approach in terms of real convergence assessment is related to the Optimal Currency Area (OCA) theory, introduced by Mundell (1961).

From the OCA theory point of view, when a country wants to join in a monetary union there should be taken into consideration criteria like the convergence of economic structures, business cycle synchronization, demand and supply shocks correlation, labor market and market flexibility in general, level of economic openness etc. We study the business cycle synchronization and demand and supply shocks correlation which are important criteria when we assess the costs of loosing the monetary policy independence and the exchange rate as an adjustment mechanism. Objectives of the paper investigate business cycles of the New Member States economy and their symmetry to the Euro Area economy using the correlation of business cycle approach, based on consensus estimations for the output gap and Principal Components Analysis (PCA) identify the demand and supply shocks using the structural vector autoregression (SVAR) and to look if there is symmetry between shocks of the New Member States and Euro Area to use various criteria, namely the Maastricht criteria, in order to group different countries in clusters, based on a clustering algorithm Optimum currency area literature review Within OCA theory authors emphasize various criteria: Production factors mobility, especially labor force (Mundell, 1961) - can reduce the need to adjust real factor prices, and the nominal exchange rate, between countries in response to disturbances. The level of economic openness (McKinnon, 1963, Alesina and Barro, 2002). Production and consumption diversification (Kenen, 1969, Tavlas, 1993). diminishes the possible impact of shocks specific to any particular sector. Wage and price flexibility (Friedman, 1953) - the transition towards adjustment following a shock is less likely to be associated with sustained unemployment in one country and/or inflation in another. Business cycle synchronization and demand and supply shocks symmetry (Cohen and Wyplosz, 1989; Weber, 1990; European Commission, 1990) risks for an asymmetric shock.

Fiscal policy integration (Kenen, 1969) and political integration (Mintz, 1970). Financial markets integration (Ingram, 1962). Financial market integration can reduce the need for exchange rate adjustment. Inflation differential (Fleming, 1971). Similarities of inflation rates are also needed to create an OCA. Optimum currency area and clustering analysis literature review Correlation of business cycles - Boone and Maurel (1998), Fidrmuc (2001), Darvas and Szapary (2005). In the field of demand and supply shocks synchronization an important contribution came from Bayoumi and Eichengreen (1992) when they recovered the underlying demand and supply disturbances using the technique developed by Blanchard and Quah (1989). Fidrmuc and Korhonen (2001), Frenkel and Nickel (2002), Horvath and Ratfai (2004) conclude that the correlation of shocks varies considerably between eurozone and accession countries. For the third objective of our paper Artis and Zhang (1998), Boreiko (2003) used the both the convergence criteria by applying a clustering algorithm to find similarities between the countries of European Union. Models used and estimation methodology Output gap estimation and the Consensus measure To obtain a measure of business cycles we used the output gap extracted with the help of four univariate methods: Quadratic trend (QT), HodrickPrescott filter (HP), Band-Pass filter (BP) and Wavelet transformation. As each filtering techniques has advantages and some weaknesses, we adopted a method similar to Darvas and Vadas (2005), in order to obtain a consensus measure of the output gap. A method is considered better if it leads to smaller revisions of past inference as new information arrived. The method gave weights to the output gaps estimated by the four filters proportional to the inverse of average revision for each method.

Principal Components Analysis (PCA) -the output gap is calculated like a linear combination of the series in the group with weights given by the first eigenvector of the first component principal . Consensus The size of revision at a certain date is: ( m) k t( m ) ( m) (m) ( m) _ _ _ _ 1 T 1 T (q t , s q t , s 1 ) ( q t q t , s ) ( q t q t , s 1 ) lt s k 1 l t s k 1 The average revision for the method is: k ( m) 1 T 1 (m) kt T 1 t 1 The weights that will be used are: 1

(m) k m p j 1 1k ( j ) Consensus Ouput Gap: _ p _ (m) c t j c j 1 t ,T Supply and demand shocks identification The model underlying the methodology of recovering shocks is the Aggregate Supply and Demand model. Methodology to extract the shocks: Blanchard and Quah (1989) and Bayoumi and Eichengreen (1992). Bayoumi and Eichengreen (1992) estimated a VAR with two variables: differences of GDP and prices level. To identify the structural shocks they imposed the long run restriction that demand shocks do not affect the level of output, but have a permanent impact on prices while supply shocks permanently affect both output and prices. Clustering analysis Kmeans clustering algorithm -used to identify a number of homogenous clusters in which we could include the twelve countries that entered in European Union after 2004 ; The data set consists of n objects (countries) with p variables (various criteria). Each variable is standardized with zero mean and standard deviation one in order to treat them as having equal importance in determining the structure. It finds a partition in which objects within each cluster are as close to each other as possible, and as far from objects in other clusters as possible.

The dissimilarity coefficient between two objects is defined as a Euclidean distance: p d (i, j ) (x ki x kj ) 2 k 1 The center for each cluster is the point to which the sum of distances from all objects in that cluster is minimized. Clustering algorithm In order to analyze how well the data is partitioned, we used a set of statistics named the silhouette width for each object and average silhouette for total dataset. b(i ) a(i ) s (i ) max(a(i ), b(i )) where a(i) is defined as average dissimilarity of object i to all objects in the same cluster and b(i) as the minimum across all other clusters of average dissimilarity of object i to all objects in each cluster. When s(i) is close to one it is implied that the object is well classified into an appropriate cluster. A value near to zero indicates the ambiguity in deciding to which cluster the object might belong. Negative values indicate that the object is misclassified. Data

Sample: 1997Q1-2009Q1, 49 observations Source of data: Eurostat and NIS Countries included in the analysis: We used eleven countries that entered in EU after 2004 and 2007(Poland, Czech Republic, Slovakia, Hungary, Slovenia, Estonia, Latvia, Lithuania, Cyprus, Bulgaria and Romania), nine countries from the eurozone (Belgium, Germany, Ireland, Spain, France, Italy, Netherlands, Austria, and Portugal), two EU member countries outside eurozone (Sweden and United Kingdom). Variable for the first objective: -quarterly GDP series (NSA) in constant prices (2000=100) ; -Tramo/Seats procedure to adjust seasonally ; Data For the structural VAR model with two variables, we used real GDP indices and the GDP deflator. Nominal GDP (NSA) and Real GDP (NSA) from Eurostat and NIS Inflation: GDP Deflator = (Nominal GDP) / (Real GDP) * 100 For the structural VAR with three variables, we used the two variables from above and the real effective exchange rate from Bank of International Settlements. For algorithm of clustering we used the nominal convergence criteria (without exchange rate criteria because a part of the countries adopted already euro) for all twelve NMS that entered in EU after 2004. Empirical results- Results for business cycle correlation the output gap based on Quadratic trend, Hodrick Prescott filter and Band Pass filter using program codes in Eviews 5.1. For the Wavelet transformation we wrote a procedure in Matlab 7.1. For the quadratic trend we estimated a regression where the cycle is the residual of a regression. For Hodrick Prescott filter we used the parameter equal with 1,600. We used for Band Pass filter the asymmetric Christiano Fitzgerald approximation with stationarity assumption of I(1) unit root process ( we tested before the stationarity). We imported the cycles obtained by Wavelet transformation from Matlab in Eviews.

We started our recursive estimation in 2001Q1. We have 29 estimation of output gaps for each country, after that we calculate the revisions, the average revisions, the weights and the consensus. Revisions of business cycles of Romania 12 ro QT 8 12 ro HP 8 4 0 0 -2 0 -4 ro BP -4 -4 -8 -12

-6 -8 97 98 99 00 01 02 03 04 05 06 07 08 3 4 2 4 4 6 ro WT 2 -8 97 98 99 00 01 02 03 04 05 06 07 08 10 8 ro PCA 6 1 -3 -2 -4 -4

-5 -2 -4 -6 97 98 99 00 01 02 03 04 05 06 07 08 Consensus 0 0 -2 ro 2 2 -1 6 4 4 0 97 98 99 00 01 02 03 04 05 06 07 08 8 -6 97 98 99 00 01 02 03 04 05 06 07 08 97 98 99 00 01 02 03 04 05 06 07 08 Revisions of business cycle of eurozone 4

3 eur QT 3 2 2 1 1 0 0 -1 -1 -2 -2 -3 -3 -4 -4 HP eur WT 1.5 1.0

3 BP 1 0 -1 -2 97 98 99 00 01 02 03 04 05 06 07 08 3 eur 2 -5 97 98 99 00 01 02 03 04 05 06 07 08 2.0 eur eur PCA 97 98 99 00 01 02 03 04 05 06 07 08 3 2 2 1 1 0 0

-1 -1 eur Consensus 0.5 0.0 -0.5 -1.0 -2 97 98 99 00 01 02 03 04 05 06 07 08 -2 97 98 99 00 01 02 03 04 05 06 07 08 97 98 99 00 01 02 03 04 05 06 07 08 Revisions and weights Revisions BG CZ EE LV LT HU PL RO SI SK

CY EUR NMS avg QT 0.14 0.18 0.49 0.50 0.41 0.16 0.14 0.30 0.21 0.20 0.09 0.12 0.24 HP 0.11 0.15 0.38

0.41 0.28 0.15 0.09 0.16 0.18 0.20 0.05 0.10 0.19 BP 0.17 0.13 0.35 0.34 0.33 0.15 0.07 0.19 0.23 0.24 0.04

0.12 0.20 WT 0.32 0.18 0.50 0.62 0.34 0.17 0.17 0.21 0.23 0.27 0.13 0.12 0.27 Weights based on principal components QT 0.34 0.26 0.26 0.26

0.25 0.27 0.27 0.27 0.26 0.26 0.29 0.27 0.27 HP 0.35 0.28 0.28 0.28 0.27 0.29 0.28 0.29 0.26 0.27 0.29

0.27 0.29 BP 0.29 0.24 0.26 0.26 0.25 0.26 0.28 0.22 0.24 0.24 0.29 0.26 0.26 WT 0.01 0.22 0.21 0.20 0.24

0.18 0.16 0.22 0.23 0.23 0.13 0.20 0.19 Weights based on revisions of percentage point output gaps - Consensus QT 0.29 0.22 0.21 0.22 0.20 0.25 0.19 0.17 0.25 0.28 0.18 0.24

0.23 HP 0.36 0.27 0.28 0.27 0.30 0.27 0.30 0.32 0.30 0.27 0.32 0.29 0.29 BP 0.23 0.29 0.30 0.33 0.25

0.26 0.36 0.26 0.23 0.23 0.37 0.23 0.28 WT 0.12 0.22 0.21 0.18 0.25 0.22 0.15 0.24 0.23 0.21 0.13 0.24 0.20

Correlation of business cycles of the NMS with eurozone(Consensus) Correlation of bussiness cycles of NMS w ith euro area(1997q1-2009q1) Romania Bulgaria Lithuania Slovakia Poland Hungary Estonia Latvia Cyprus Czech Republic Slovenia 0.80 0.60 0.40 0.20 0.00 Correlation of business cycles of NMS with euro-area 1997q1-2001q4 2002q1-2009q1 Average 2002q1-2008q2 0.95 2002q1-2008q2

Average 1997q1-2001q4 Average 2002q1-2009q1 0.75 0.55 0.35 H ungary R om ania P oland B ulgaria Latvia Estonia C yprus Lithuania -0.45 C zech R epublic -0.25 S lovakia -0.05 S lovenia 0.15 We split the sample in two periods to see how the correlation evolved in time.

We found that correlation increased substantially in the last years for all the NMS. Because we suspect that the recent economic crisis had an impact on the correlation of business cycles between NMS and eurozone, we computed the correlations excluding the last 3 quarters from the sample .On average the correlation coefficients for 2002q1-2008q2 are smaller than 2002q1-2009q1 Correlation of business cycles of the eurozone members with eurozone We checked also if there is an endogeneity of the OCA, ie. the creation or the joining to European Monetary Union has impact on business cycles correlation in the sense of increasing or decreasing it. Correlation of business cycle of the eurozone members with eurozone: 1.2 1997:1-2001:4 1 2002:1-2009:1 0.8 0.6 0.4 -0.6 Ire la n d P o rtu g a l C y p ru s S lo v a k ia S p a in F ra n c e

A u s tria B e lg iu m S lo v e n ia -0.4 N e th e rla n d s -0.2 Ita ly 0 G e rm a n y 0.2 Contemporaneous correlation The Optimum Currency Area (OCA) theory affirms that the business cycle correlation should be positive, strong and contemporaneous between a NMS and eurozone. In order to check the lag for which the correlation coefficient is the largest, we estimated the correlation of business cycle with some lags. Correlation of business cycle between NMS (at time t+i) and eurozone (at time t) for 2002q1-2009q1 period: Correlation of business cycles of the NMS with eurozone Correlation of business cycle between NMS (at time t+i) and eurozone (at time t) for 1997q1-2009q1 period how much the business cycles of a NMS should be correlated with the business cycle eurozone in order to have net benefits from euro adoption? Artis (2004) argues that the literature doesnt help us too much but probably the best criteria will be that the NMS country which wants to adopt euro

should not have smaller synchronization with eurozone than the existing members of eurozone does. Fidrmuc and Korhonen (2006) said that if business cycle correlation in a new EU member state is higher than correlation of a peripheral euro area economy (e.g. Ireland , Portugal) we have confidence that the NMS has progressed far enough in fulfilling this OCA criterion. Correlation of business cycles of EU members with eurozone: 1.00 1997q1-2001q4 0.75 2002q1-2009q1 0.50 0.25 0.00 U n ited K in d o m H u n g ary Ro m an ia Po lan d in the case of the other old Bu lg aria Irelan d L ith u an ia Po rtu g al E sto n ia L atv ia Cy p ru s

Sw ed en Czech Rep u b lic Slo v ak ia Sp ain Fran ce A u stria Belg iu m Slo v en ia Italy -0.50 G erm an y members of EMU. N eth erlan d s -0.25 For the recent period, Slovenia, Slovakia and Czech Republic has even a higher correlation with eurozone than Results for demand and supply shocks correlation For the structural VAR decomposition we worked with real GDP growth series and inflation rate. We used the Akaike criterion to select the optimal lag length for the VAR. After we obtained the form of VAR we imposed the structural restriction suggested by Bayoumi and Eichengreen (1992), We defined a matrix of restrictions in Eviews with two lines and two columns that contain the structural condition and that matrix pass on the effects of shocks to the two variables growth GDP rates and inflation rate. We made the residuals (shocks) for each country and then we studied the similarity of shocks by doing the correlation between demand shocks of a NMS and eurozone and afterwards the correlation between supply shocks.

Correlation of demand (OY axis) and supply shocks (OX axis), 1997-2009 1 Euro zone 0.8 IT 0.6 0.4 ES FR DE NL 0.2 AT IE PT LT CY 0 SK PL UK CZ BG -0.2 SI SE

HU LV BE RO EE -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 Correlation of demand (OY axis) and supply shocks (OX axis), 2002-2009 1 Euro zone 0.8 IT DE 0.6 NL FR 0.4 ES HU

LV IE 0.2 LT SK PL 0 PT UK CY EE SE SI RO AT CZ -0.2 BE BG -0.4 0 0.2 0.4 0.6 0.8 1

Supply (left) and demand (right) shocks for Romania and eurozone, 2002-2009 .03 .03 .02 .02 .01 .01 .00 .00 -.01 -.01 -.02 -.02 -.03 -.04 -.03 2002 2003 2004 2005 Romania 2006 2007

Eurozone 2008 2002 2003 2004 2005 Romania 2006 2007 Eurozone 2008 Structural VAR model with three variables GDP growth rate, inflation rate and real exchange rate growth rate. Three types of shocks exert influence in this specification of VAR: supply, real demand (IS), and nominal demand shocks. This is based on an assumption that GDP growth rate (supply) is in the longterm horizon independent of both real exchange rate and inflation rate. Real exchange rate growth rate (real demand) may in the long-term horizon depend on GDP growth rate, but it is independent of inflation rate. Inflation rate (nominal demand) may depend on both GDP growth rate and real appreciation rate. Correlation of shocks between Romania and eurozone: Results for clustering analysis We used again the Matlab 7.1 package to perform a Kmeans clustering algorithm to find homogenous groups in data.

Allocation of countries by clusters: If we look at the value of average silhouette we can say that from statistical point of view, the data was well partitioned. Conclusions (I) This paper assesses the degree of readiness of New Member States (NMS) of European Union, including Romania, to adopt euro, mainly based on optimal currency area (OCA) criteria. Using a consensus measure of output gap computed from 4 filtering techniques plus a benchmark method based on Principal Component Analysis (PCA), we estimated the business cycle correlation between NMS and eurozone. Our findings suggest that the correlation of the business cycle in the case of Romania is one of the lowest among NMS, followed by Bulgaria, Slovakia and Lithuania although increased tremendously in the last years. Our results suggest also that the financial and economic crisis which hit the world economy recently led to an increase in the business cycle correlation between NMS and eurozone, as the countries are simultaneously faced with a sharp GDP contraction. According with our results, for the most recent sample (2002-2009), the business cycle of the NMS countries outside the eurozone is contemporaneously and positive correlated with eurozone. Conclusions (II) For the most of the NMS the correlations of demand shocks with eurozone are negative, excepting Slovenia and the correlations of supply shocks with eurozone are positive except Cyprus. In the case of Romania, our results suggest that the supply shocks are correlated positively with eurozone and the correlation is quite high.

Nevertheless, in terms of demand shocks, the correlation is negative. We performed also a structural VAR model with three variables for Romania and eurozone: GDP growth rate, inflation rate and real exchange rate growth rate. The results show that the demand shocks for Romania are negatively correlated with eurozone for the full sample, both for real and nominal shocks. Using the Maastricht criteria for the NMS, we identified in the paper some homogenous groups of countries based on clustering analysis. According with our estimations, the optimal number of clusters for the NMS should be 4, Romania being in the same cluster with Estonia, Lithuania and Latvia. The main conclusion of our paper is the fact that Romania, as well as some other NMS countries still need time to progress on the real convergence criteria in order to adopt euro without major costs. References Afonso, A. and D. Furceri (2007), "Sectoral Business Cycle Synchronization in the European Union", Department of Economics at the School of Economics and Management (ISEG), Technical University of Lisbon, Working Papers No. 02. Alesina, A., R. J. Barro, and S. Tenreyro (2002), "Optimal Currency Areas", National Bureau of Economic Research Working Papers, 9072 Artis, M. and Zhang, W.(1998), Membership of EMU: A Fuzzy Clustering Analysis of Alternative Criteria, EUI Working Paper 98/52. Artis, M., M. Marcellino and T. Proietti (2004) Characterising the Business Cycle for the Accession Countries, Innocenzo Gasparini Institute for Economic Research (IGIER), working paper no. 261 Bayoumi, T. and B. Eichengreen (1992), Shocking Aspects of Monetary Unification, NBER Working Paper No.3949 Blanchard, O. and D. Quah (1989), The dynamic effects of aggregated demand and supply disturbances, American Economic Review, Vol. 79, No. 4, pp. 655-673

Boone, L. and M. Maurel, (1998), Economic Convergence of the CEECs with the EU, Centre for Economic and Policy Research Discussion Paper, no. 2018, 1998 Boreiko, D. (2002), "EMU and Accession Countries: Fuzzy Cluster Analysis of Membership", sterreichische Nationalbank Working Paper, No. 71. Canova, Fabio, 1998. Detrending and Business Cycle Facts. Journal of Monetary Economics 41, 475512. Cohen, D. and C. Wyplosz (1989), "European Monetary Union: An Agnostic Evaluation", CEPR Discussion Paper Series, No. 306. Conway, Paul and D. Frame (2000): A spectral analysis of New Zealand output gaps using Fourier and wavelet techniques, paper presented at Reserve Bank of New Zealand. Darvas, Z. and G. Szapry (2004),Business Cycle Synchronization in the Enlarged EU: Comovements in the New and Old Members, Magyar Nemzeti Bank Darvas, Z. and G. Vadas (2005), A New Method for Combining Detrending Techniques with Application to Business Cycle Syncronization of the New EU Members, Magyar Nemzeti Bank References DeGrauwe, P. (1997), The Economics of Monetary Integration, Oxford: Oxford University Press. Enders, W. (2004), Applied Econometric Time Series 2nd ed. Wiley. European Commission (1990), One Market, One Money. An Evaluation of the Potential Benefits and Costs of Forming an Economic and Monetary Union, European Economy, No. 44. Fidrmuc, J. and I. Korhonen (2006), Meta-Analysis of the Business Cycle Correlation between the Euro Area and the CEECs, CESifo Working Paper No. 1693 Fidrmuc, J., Korhonen, I. (2001), Similarity of Supply and Demand Shocks between the Euro Area and the CEECs, BOFIT Discussion Paper 13, Bank of Finland, Institute for Economies in Transition, Helsinki Fleming, M. J. (1971), "On Exchange Rate Unification", Economic Journal, 81(323): 467-88. Frankel, J. A., Rose A.K. (1996), The Endogeneity of the Optimum Currency Area Criteria, CEPR Discussion Paper No. 1473 Frenkel, M. and Nickel, C. (2002), How symmetric are the shocks and the shock adjustment dynamics between the Euro Area and Central Eastern European Countries?, International Monetary Fund, Working

Paper, No. 222 Friedman, M. (1953), The Case for Flexible Exchange Rates. Essays in Positive Economics, Chicago: University of Chicago Press. Horvath, J. and A. Ratfai (2004), "Supply and demand shocks in accession countries to the European Monetary Union," Journal of Comparative Economics, Vol 32, No.2, 202-211 Ingram, J. C. (1962), Regional Payment Mechanisms: The Case of Puerto Rico, Chapel Hill: University of North Carolina Press. Kenen, P. (1969) The Theory of optimum Currency Areas: An Eclectic View, in Mundell, R. and A. K. Swoboda (eds.) Monetary Problems of the International Economy, University of Chicago Press, 41-60 References McKinnon, R. (1963) Optimum Currency Areas, American Economic Review, vol. 52, 717-725 Mintz, N. N. (1970), "Monetary Union and Economic Integration", The Bulletin of New York University Graduate School of Business Administration (Institute of Finance), No. 64. Mongelli, F. P. (2002): NewViews on the Optimum Currency Area Theory: What is EMU Telling Us? European Central Bank Working Paper, no. 138, April 2002. Mundell, R. A. (1961), "A Theory of Optimum Currency Areas," American Economic Review 51. NBS Research Department (2006), The effects of euro adoption on the slovak economy, Paper presented at National Bank of Slovakia. Obstfeld, M. Krugman, P.R. (1994), International Economics, Theory and Policy. New York, Harper Collins College Publishers 1994. Rinaldi-Larribe, M.-J. (2008),Is economic convergence in New Member States sufficient for an adoption of the Euro?, The European Journal of Comparative Economics, Vol. 5, n.2, pp. 133-154. Weber, A. (1990), "EMU and Asymmetries and Adjustment Problems in the EMS: Some Empirical Evidence", CEPR Discussion Paper Series, No. 448.