What drives EMU current accounts? A time varying structural VAR approach. Maximilian Podstawski

Transcription

1 What drives EMU s? A time varying structural VAR approach Maximilian Podstawski Abstract This paper investigates the drivers of the divergent balances that emerged in the EMU in the run-up to the most recent crisis. A time-varying structural VAR model identified via long-run and sign restrictions is set up in order to integrate the partly competing hypotheses discussed in the literature into a single structural framework. Besides excessive domestic demand, mismatched monetary policy and shocks to price competitiveness have been driving the divergence during the first ten years of EMU, whereas supply side effects are negligible. The rapid rebalancing within EMU is found to be attributable mainly to demand side adjustment, while in addition the adverse effects of deteriorated price competitiveness have diminished. Keywords: Current account; monetary union; historical decomposition; sign restrictions. JEL classification: E00, F32, F4. I am thankful to Helmut Lütkepohl, Simon Junker, Dieter Nautz and seminar participants of the Empirical Macroeconomics Workshop 2013/14 at Freie Universität Berlin and of the DIW Graduate Center Winter Workshop 2013/14 for useful discussions and helpful comments to earlier versions of this paper. All remaining errors are my own. German Institute for Economic Research (DIW Berlin Graduate Center) and Freie Universität Berlin; mpodstawski@diw.de

2 1 Introduction Significant macroeconomic imbalances have evolved in the European Monetary Union (EMU) in the run-up to the most recent crisis. They became particularly apparent in the form of steadily diverging s within the monetary union, while the of the monetary union as a whole remained rather unchanged. By now there has been considerable adjustment underway in those countries that had build up significantly negative net foreign asset positions and were at the core of the crisis. Various authors have pointed towards private indebtedness and persistent deficits with their impact on the stability of the countries banking systems and solvency being at the roots of the recent crisis. Barrios et al. (2009) find that large deficits amplify the impact of deteriorated public finances on government bond spreads, while the IMF (2010) concludes that deficits are correlated with higher sovereign credit default swap (CDS) spreads. Gros (2011) argues that foreign debt, i.e. accumulated deficits, form the underlying problem for the solvency of the eurozone countries. Lane and Pels (2012) see the imbalances at the core of the current crisis, having contributed to the extent of the economic contraction during the crisis and severely damaged the banking system. This has triggered an intense debate in academic and policy circles on the drivers of the divergence and, hence, the appropriate policy response. Given that the widening of current accounts coincided with the launch of the monetary union, it seems obvious to relate this phenomenon to changes in the institutional setting of the EMU member countries: the abolishment of an exchange rate mechanism, the introduction of a common monetary policy and the intensified financial integration as seen in the rapid convergence of nominal s. Although there is consensus that the financial integration through a reduction in risk premia and, thus, borrowing costs within EMU is underlying the pronounced divergent macroeconomic development, the channels through which it has fed into the positions of the member countries are discussed controversially in the literature. Fundamentally, the competing views attribute the divergence either to convergence on the supply side, changes in price competitiveness or to demand distortions. The earliest hypothesis understands the imbalances that emerged in the 1

3 EMU along the lines of the intertemporal theory of the (Obstfeld and Rogoff, 1995) as the result of a convergence process within the monetary union that allowed those economies that were catching up to borrow against future growth. According to this hypothesis financial integration and the elimination of exchange rate risks resulting in the convergence of nominal s fueled the convergence process and thus the growing capital flows within the monetary union reflected in the divergence (Blanchard and Giavazzi, 2002). While several authors find evidence pointing towards the validity of this hypothesis (Ca Zorzi and Rubaszek, 2008; Ahearne et al., 2008; Lane and Pels, 2012), it has been challenged by others (Barnes et al., 2010; Holinski et al., 2012). A second strand of the literature views the development of price competitiveness at the roots of the divergence (Arghyrou and Chortareas, 2008). With the introduction of the common currency two main (external) adjustment mechanisms, the flexible exchange rate and an autonomous monetary policy came to an end. Shifts in relative prices, triggered by idiosyncratic shocks or the heterogeneous reaction to one common shock in absence of sufficient adjustment mechanisms lie according to this view at the root of the diverging s. Belke and Dreger (2011) find evidence pointing towards movements in relative prices, driven by excessive nominal wage growth in deficit countries, being at the center of the divergence. A different line of reasoning suggests that the imbalances are part of an overall macroeconomic divergence due to the mode of operation of the so called real channel. Walters (1990) argued with reference to the European Exchange Rate System (ERM), that fixed nominal exchange rates and liberalized capital markets within a group of countries that is heterogeneous with respect to their inflation rates would create an inherently unstable system. Inflation differentials given a convergence of nominal s would materialize in real differentials. Mongelli and Wyplosz (2008) extend this line of reasoning by taking into account effects stemming from movements of relative international prices. The demand reaction triggered by real differentials feeds into domestic prices but also into relative international prices. Therefore, countries with relatively high inflation rates within a monetary union will experience rising domestic demand due to low real s but also a deterioration of their price competitiveness. The rise in demand on the one hand and the shift of relative prices in favor of foreign goods on the other hand boosts the demand for foreign 2

4 goods, stabilizing the domestic price level at the cost of a deterioration of the. Following this line of arguments, the divergence can be either traced back to mismatched monetary policy (Wyplosz, 2010) or domestic demand forces potentially amplified by the real channel (Wyplosz, 2013). All of the hypotheses sketched above are very much in line with the stylized facts 1 presented in Figure 1: While the divergence was clearly driven by private sector developments rather than public borrowing or dissaving, it was accompanied by differentials in real s that switched signs at the entry into EMU, persistent GDP growth differentials and diverging price competitiveness. However, simultaneity does not imply causality and the hypotheses differ severely with regards to the underlying drivers and, hence, to the policy conclusions for overcoming existing and preventing the build-up of future macroeconomic imbalances within EMU. So far and partly due to the small number of observations the empirical literature has focused on exploring correlations in reduced form frameworks (see Jaumotte and Sodsriwiboon (2010), Barnes (2010), Belke and Dreger (2011), Lane and Pels (2012), Schnabl and Wollmershäuser (2013) and Atoyan et al. (2013) among others). The aim of the paper is to integrate the partly competing hypotheses regarding the divergence of EMU s into one structural framework in order to decompose the into contributions from potential drivers and assess their validity while circumventing the EMU small sample problem. In addition to that, the framework allows to investigate the source of the pronounced recent adjustment in EMU s as well as potential changes to the transmission and propagation of shocks related to the entry into EMU. The analysis is based on a time varying structural vector autoregressive (VAR) model. The shocks of interest are identified via a combination of long-run and sign restrictions. It is found that, depending on the country considered, excessive demand and over-accommodating monetary policy in addition to movements in relative prices account for a substantial fraction of the variance in EMU s. However, convergence processes seem to have played 1 Figure 1 plots GDP weighted averages of the two clusters. Clustering has been undertaken based on the position relative to GDP from 1999 to 2008 applying the k-means algorithm. The periphery cluster consists of Greece, Ireland, Italy, Portugal, Slovenia, Slovakia, Spain. The core cluster consists of Austria, Belgium, Finland, France, Germany, Luxembourg, Netherlands. See Holinski et al. (2012) for a similar approach. 3

5 NEWU SEWU NEWU SEWU NEWU SEWU (a) by GDP (b) net savings, private sector (c) net savings, public sector NEWU SEWU NEWU SEWU NEWU SEWU (d) real. s (e) GDP, growth rate (f) Inflation Figure 1: Selected macro variables by cluster (solid line = core cluster, dashed line = periphery cluster) with shaded areas indicating the period of divergence during the first ten years of EMU, Source: Ameco and OECD. a minor role for the divergence of the s of the considered EMU member states. The recent adjustment that has taken place within the EMU on the side of the former deficit countries is almost solely due to contracting demand and a recovering price competitiveness, while no adjustment so far has taken place in Germany. The paper is organized as follows. In Section 2, we set up a time varying structural VAR model with an exogenously set break date at the entry into monetary union and introduce the data. In Section 3, we derive the long-run and sign restrictions for the identification of the four relevant shocks, a, a demand shock, a supply shock, and a. Section 4 introduces the data used, while Section 5 discusses the impact of the identified structural shocks on macro variables and the historical evolution of the for a sample of Euroarea countries. Section 5 concludes. 4

6 2 TV-VAR In this section, a time varying VAR model with an exogenous break at the time of entry into the monetary union is set up. The 4 1 vector Y t contains the log difference of a GDP π t, log difference of real GDP y t, the by GDP ca t and a short term interest rate r t. The reduced form VAR model is given by Y t = ψd t + A 1 Y t A p Y t p + u t, t = 1,..., T, where u t is a 4 dimensional White Noise process with positive definite covariance matrix Σ u, A j for j = 1,..., p are the 4 4 coefficient matrices, p is the lag order of the VAR, D t is a vector capturing the deterministic components of the model, ψ t is the matrix of respective coefficients and T is the sample length. Written more compactly using Π := [A, ψ], where A := [A 1,..., A p ] and Z t := (D t, Y t 1,..., Y t p) gives Y t = ΠZ t + u t, with E(u t u t) = Σ u. Now, consider a structural change in the model occurring at T B, in our case in the last period before the introduction of the Euro, so that Y t = Π(t)Z t + u t, E(u t u t) = Σ u (t), where Π(t) := Π 1 1(t T B ) + Π 2 1(t > T B ) Σ u (t) := Σ u,1 1(t T B ) + Σ u,2 1(t > T B ), with Π 1 := [A 1, ψ 1 ] and Π 2 := [A 2, ψ 2 ] and 1( ) is an indicator function. Given that data on s and GDP is not available at any frequency higher than quarterly, there are potentially many parameters to be estimated based on the relatively short EMU sample. Therefore, the model will accommodate the potential regime change at the 5

7 entry into monetary union, while keeping it as parsimonious as possible in terms of time variance. A battery of results from different Chow tests is reported in Section 4 and supports the specification of the model. So far the reduced form model is put into place. The following section will proceed by bringing non-sample information into play in order to identify economically meaningful structural shocks, namely a, a, a domestic supply shock and a. 3 Identification The structural VAR system associated with the reduced form is given by u t = B(t)ɛ t, where ɛ t are the structural shocks obtained by a linear transformation of the reduced form errors u t and B(t) := B 1 1(t T B ) + B 2 1(t > T B ), that is, the structural parameter matrix B, is allowed to be time varying with respect to the second regime. Normalizing the variances of the structural innovations for both regimes to one, i.e. assuming E(ɛ t ɛ t) = I K gives Σ u (t) = E [u t u t] = E [B(t)ɛ t ɛ tb(t) ] = B(t)B(t) In order to integrate the partly competing hypotheses regarding the sources of fluctuations within the monetary union outlined above, there is the need to identify four structural shocks to the economy summarized by the VAR system: a, ɛ ds t, to capture potential convergence, a, ɛ comp t, to account for exogenous movements in relative prices driving the s, a, ɛ dd t, and a, ɛ m t, that capture potential demand distortions in the considered economies. The identification approach follows Canova and Nicoló (2002) and Uhlig (2005) by imposing sign restrictions on the impulse responses of selected variables for a limited time period after the 6

8 shock occurrence. In the following, shocks are identified by imposing restrictions on the variables for four quarters after the shock occurred. Table 1 summarizes the sign restrictions imposed to identify the four respective shocks: A within a monetary union leads to an increase in inflation that is associated with no decrease of s. A domestic demand shock is defined as increasing the GDP growth rate and inflation while no reduction of s occurs. A impacts inflation negatively, GDP growth positively and, again, does not involve a decrease of s. Finally, an expansionary is defined as leading to a reduction of s accompanied by an increase in inflation and GDP growth. While leaving the response of the variable unrestricted in order to let the data speak as this is the variable at the center of the following analysis, the restrictions on the reaction of inflation, GDP and the reflect those backed by a broad range of models. Table 1: Sign restrictions for impulse responses ɛ comp t ɛ dd t ɛ ds t inflation π t gross domestic product y t ca t r t ɛ m t However, imposing the sign restrictions is not sufficient for the set of shocks to be properly identified, since the can not be disentangled from the domestic demand shock based on the sign restrictions defined in Table 1. Additional identifying information comes into play by imposing a single long-run restriction (Blanchard and Quah, 1989), i.e. forcing the response of GDP to a to be zero in the long-run as indicated in Table 2. Table 2: Long run restrictions for impulse responses ɛ comp t ɛ dd t ɛ ds t inflation π t gross domestic product y t 0 ca t r t ɛ m t 7

9 Intuitively speaking, the restrictions are implemented as follows: Various structural impact parameter matrices B(t) that fulfill the single zero long-run restriction are considered and a matrix is dismissed if the implied impulse responses do not match the defined sign restrictions. More formally, the combination of long-run and sign restrictions is implemented applying an algorithm similar to the one proposed by Binning (2013) for combining short- and long-run restrictions with sign restrictions in underidentified models. An initial matrix of structural parameters B determined by a Choleski decomposition of Σ u is estimated, assuring the orthogonality of the structural shocks, so that u t = Bɛ t where Σ u = BB. In a second step the matrix of structural parameters B is rotated randomly. Let therefore N = QR be the QR-decomposition of an independently drawn standard normal K K matrix N, where Q is an orthogonal rotation matrix, i.e. QQ = I (Rubio-Ramirez et al., 2010). The rotation matrix Q is post-multiplied to the initial matrix of structural parameters B. Now B = BQ is a random rotation of the initial structural parameter matrix, where Σ u = B B = B QQ }{{} B. =I Thirdly, the zero restrictions have to be imposed on the long-run impact matrix Ξ = (I K A 1... A p ) 1 B. In order to achieve this, the long-run impact matrix Ξ is rotated such that the long-run restrictions are met. This is done by post-multiplying an appropriately defined Givens rotation matrix 2 G to the long-run impact matrix, i.e. Ξ = Ξ G, where Ξ now satisfies the imposed long-run restrictions. In a final step, the structural parameter matrix B = BG = BQG is used for the construction of impulse responses and the model is dismissed if the specified sign restrictions are not met. These steps are repeated until 1000 models matching the sign restrictions are drawn. Following Barnett and Straub (2008) in order to take into account both, the estimation uncertainty of the reduced form model and the identification uncertainty, the model is bootstrapped (Benkwitz et al., 2001) and the covariance matrix Σ u is reestimated after each draw. 2 The Givens matrices assuring that the zero restriction of the model under consideration are met is constructed via Algorithm in Golub and Van Loan (2012). 8

10 Since the system with one long-run restriction and a number of sign restrictions remains underidentified, resulting in a set of models satisfying the specified restrictions, the problem arises of how to summarize the information contained in the impulse responses. A number of strategies to handle this issue have been suggested up to now. Uhlig (2005) makes use of pointwise median impulse responses in order to capture the median tendency of the impulse responses. However, those pointwise medians may represent information about shocks that stem from different models and there is no guarantee that these shocks will be uncorrelated. This would be especially problematic for historical decompositions as conducted in Section 5. Therefore, this paper resorts to the suggestion by Fry and Pagan (2011) of using closest to median impulse responses. The idea behind the concept is to choose the model with the impulse responses closest to the median impulse responses as representative for the set of models. By doing so, it is assured that impulse responses are produced by a single model and thus are consistent, while retaining the notion of median measures being an appropriate summary of the set of models. 4 Model specification In order to bring the model to the data, it has to be specified in terms of the time variation of the parameters and the lag structure. The set of countries under consideration consists of four EMU member states running deficits Spain, Italy, Portugal and Ireland on the one hand and Germany, running surpluses in the aftermath of the launch of EMU, on the other hand. The exogenous break date is set to be 1998Q4, which is the last quarter in the pre-ewu regime, while the sample starts in 1990Q1 (1986Q1 for Spain) and ends in 2013Q4. Data on GDP s, GDP, the and short term s (3 months) at quarterly frequency are taken from the OECD Main Economic Indicators database. The first two variables enter in log differences, the latter two in levels. See figure 4 for a descriptive plot of the data. The lag length of the VAR model is set according to the modal lag length chosen by the Akaike, the Schwarz (Bayesian) and Hannan-Quinn information criterion for models with a lag length up to 12. Based on this, a VAR(2) is fitted to Spanish data, while all other models are chosen to be VAR(1). 9

11 The model described in Section 2 potentially allows for full variability of the parameters over the two exogenously identified regimes. However, given the length of time series data for the monetary union regime, it seems desirable to keep the model as parsimonious as possible in terms of time variation. In order to isolate a specification that accommodates the potential structural break between the two regimes on the one hand, while making use of the entire sample information for the estimation of as many parameters as possible, a number of Chow type tests of different model specifications is conducted. Following (Candelon and Lütkepohl, 2001), who show that Chow tests are size distorted in small samples and in particular overreject the null, bootstrapped versions of the test statistics are used for inference on the parameter stability. Table 3 reports the p-values from bootstrapped Chow type tests with the alternative of full time variability over the two regimes against different null hypotheses. Table 3: Chow type tests of the alternative of full time variability of the model null p-values from bootstrapped χ 2 ψ A Σ u Spain Italy Portugal Ireland Germany C C C 1e-07 1e TV C C 1e-07 1e C TV C 1e-07 1e-07 1e-07 1e C C TV TV TV C 1e-07 1e C TV TV TV C TV Note: C indicates constancy of the respective parameter matrix, TV indicates time variance over the regimes. Clearly, the tests indicate a break in the system at the entry into monetary union, as the null of constancy of all parameters is rejected for Spain, Italy, Portugal and Ireland. At the same time the tests indicate no break for the German system. The three specifications with a time varying covariance matrix Σ u (t) are the only specifications that can not be rejected for the majority of the country sample. The specification with constant deterministic parameters, ψ, constant parameters in the matrix of slope coefficients, A, and a change only in the reduced form covariance matrix, Σ u, is the most parsimonious one in terms of parameters to be estimated, while sufficiently accounting for the regime change. Given the short EMU time series data, this specification serves as the model of choice for the remainder of the paper. 10

12 5 Results This Section employs the outlined setup in order to assess how the monetary unification has changed the transmission and propagation of shocks in the economies and what has been driving the fluctuations in the considered EMU countries in the run-up to the most recent crisis and during the subsequent adjustment. The analysis resorts to a comparison of impulse response functions to tackle the former and to a historical decomposition of the current accounts to investigate the latter issue. 5.1 Impulse Responses Figure 2 plots the impulse responses of the four variables to a, a, a and an expansionary for the Spanish economy and both regimes. Note that the impulse responses to inflation and GDP growth are cumulated and therefore reflect responses of the GDP and real GDP and that impulse responses for both regimes, the pre-emu and the EMU regime, are plotted in order to allow for an assessment of the changes in the transmission and propagation of shocks due to monetary unification. While the impulse responses of the, GDP and the are in line with expectations in terms of sign and magnitude, attention should be drawn to the response of the variable, since it is at the center of the analysis and the only variable that remained unrestricted in its response to the four different shocks in order to let the data speak. The only shocks that impact on the significantly negatively in both regimes are the domestic demand and s, while the response of the to a is negative but insignificant. The supply shock response points towards intertemporal effects along the lines of Obstfeld and Rogoff (1995) rather than intratemporal effects dominating in the Spanish economy. The negative response of the to the expansionary is also in line with the findings of Bems et al. (2007) for the United States and might be attributable to an increase in investment and consumption activity following the reduction of s. The only shock, that does not have a significantly negative impact the in the EMU-regime is the. 11

13 Figure 2: Impulse responses for the Spanish model, 68 % confidence bands (blue=pre-emu regime, red=emu regime) The positive reaction of the to a shock to price competitiveness for the pre- EMU regime might be attributable to nominal exchange rate movements that are not accounted for in the model. The impulse response functions of the entire country sample considered in the analysis show some degree of heterogeneity especially with regards to the response to shocks. Impulse responses for the Italian, Portuguese, Irish and German economy are plotted in Figures 5 to 8 in the appendix. While most of the impulse responses seem again in line with expectations and the responses of the Spanish model, some seem to be a little less intuitive, as for example for the positive response of the Italian to a expansionary monetary policy shock. However, since the identification was set up in order to let the data speak, we will not impose any informal restrictions in terms of responses as the shocks are already identified as such by imposing restrictions elsewhere. Comparing responses in the two regimes, pre-emu and EMU, evidence clearly points towards substantial changes in the shock transmission due to monetary unification in the case of the Spanish model. In particular, monetary policy reacts less to the country specific shocks, 12

14 as one would expect given that it targets the euro area as an aggregate, while prices are also less sensitive to shocks to domestic demand and price competitiveness, potentially reflecting the integration into the single European market. While the changes of the impulse responses for the Spanish, Italian, Portuguese and Irish models are somewhat similar in terms of changes in the reaction of prices and s, the transmission of shocks has not changed significantly with the entry into EMU when it comes to the German model as indicated by Figure 8. This finding matches the results of the Chow tests (see Table 3) that did not indicate time variation in the parameters of the German VAR model in the first place and points towards the fact, that the macroeconomic environment including the monetary policy has not changed significantly as compared to the environment the German economy found itself in during the 1990s. 5.2 Historical Decomposition In order to assess the relative importance of the four different shocks and, thus, allow a judgment regarding the hypotheses outlined above, a historical decomposition of the s of all five considered countries is undertaken. The basic idea behind the concept of the historical decomposition is to differentiate between the evolution of the based on the recursive interaction of the variables in the VAR and the evolution given the structural disturbances of the model. At each point in time, the variable of interest can be decomposed into a baseline level, which would have realized in the absence of all lagged and contemporaneous shock impacts, and the contributions due to structural innovations. In other words, the baseline reflects the development an agent would have predicted based on the information set available at the starting point of the decomposition. As a note of caution it should be emphasized that the starting point of a decomposition might matter quite substantially even in the case of stationary processes. Resilient inference can only be based on periods some distance away from the starting point (Lütkepohl, 2011). Figure 3 plots historical decompositions of the for Spain, Italy Portugal, Ireland and Germany based on the estimated time varying structural VAR models. Table 4 summarizes the average contributions of the different shocks to the respective s. The historical decomposition of the variable into the contributions of structural 13

15 Figure 3: Historical Decomposition of selected EMU s including the crisis period Current_Account 0 Current_Account time time baseline price.competitiveness.shock domestic.demand.shock domestic.supply.shock monetary.policy.shock baseline price.competitiveness.shock domestic.demand.shock domestic.supply.shock monetary.policy.shock (a) Spain (b) Italy Current_Account 0 5 Current_Account time time baseline price.competitiveness.shock domestic.demand.shock domestic.supply.shock monetary.policy.shock baseline price.competitiveness.shock domestic.demand.shock domestic.supply.shock monetary.policy.shock (c) Portugal (d) Ireland 10 5 Current_Account time baseline price.competitiveness.shock domestic.demand.shock domestic.supply.shock monetary.policy.shock (e) Germany shocks points towards some heterogeneity regarding the driving forces of the divergence in the run-up to the recent crisis, but also displays some similarities in the patterns 14

16 of divergence. In all four countries running large deficits demand shocks play a major role in the phase of deteriorating s. This might be due to overoptimistic growth expectations (Lane and Pels, 2012) or other related changes in preferences and is the strongest shared pattern among these countries. The Spanish, Portuguese and Irish deficits were also fueled by expansive monetary policy after entry into EMU and subsequent demand booms. This could be an indication of the real channel being at play, given the relatively high initial inflation rates in these two countries at the onset of EMU. In addition to this, demand side effects potentially reflecting the housing bubble further deteriorated the Spanish. A significant share of Portugal s and Ireland s deficits can be attributed to deteriorated price competitiveness that does not stem from other demand side effects, while Germany s strong external position has profited mainly from improvements in price competitiveness. Supply side effects played a minor role in the evolution of the considered current accounts in the run-up to the crisis, if at all, they were stabilizing the s of the deficit countries. Table 4: Average contribution to selected EMU s (2003Q1-2010Q4) price competitiveness domestic demand domestic supply monetary policy Spain Italy Portugal Ireland Germany By now there has been considerable adjustment underway in those countries that had build up significantly negative net foreign asset positions and were at the core of the crisis. Rebalancing in those countries is heavily driven by demand side adjustments, while in addition the adverse effects of past shocks to price competitiveness have diminished in Italy and Ireland. While the deficit countries have undergone severe adjustments of their positions, this is not the case for Germany where the remains on the historically high level. Based on the findings from the historical decomposition of the s supply 15

17 side effects did not play a major role as drivers of EMU s, while there is some evidence for the role of price competitiveness especially for the Italian economy. This is not the case for Spain, Portugal and Germany. The fluctuations in these countries are to a large extent attributable to monetary policy and demand shocks. 6 Conclusion Based on a time varying structural vector autoregressive (VAR) model identified via a combination of long-run and sign restrictions it is found that, depending on the country considered, excessive demand and over-accommodating monetary policy in addition to movements in relative prices account for a substantial fraction of the variance in EMU s. However, convergence processes seem to have not played a role for the divergence of the s of the considered EMU member states. Both, s and prices become less sensitive to shocks in the periphery countries, while the shock transmission in the German economy remains unchanged at entry into EMU. In particular, monetary policy reacts less to the country specific shocks of periphery countries, as one would expect given that it targets the euro area as an aggregate, while prices are less sensitive to shocks to domestic demand and price competitiveness, potentially reflecting the integration into the single European market. The recent adjustment that has taken place within the EMU on the side of the former deficit countries is almost solely due to contracting demand and diminishing adverse effects from shocks to price competitiveness on the, while no adjustment so far has taken place in Germany. The findings emphasize the role of demand and monetary policy for the divergence makes a case for the importance of institutional reforms such as a banking union and fiscal policy integration in order to carefully watch demand dynamics in EMU and to overcome existing as well as prevent future macroeconomic imbalances and crisis-laden adjustment. 16

18 References Ahearne, A., B. Schmitz, and J. von Hagen (2008). Current account imbalances in the euro area. Challenges of globalization: imbalances and growth, 41. Arghyrou, M. and G. Chortareas (2008). Current account imbalances and real exchange rates in the euro area. Review of International Economics 16 (4), Atoyan, R., J. Manning, and J. Rahman (2013). Rebalancing: Evidence from adjustment in Europe. (74). Barnes, S. (2010). Resolving and avoiding unsustainable imbalances in the euro area. OECD Economics Department Working Papers (827). Barnes, S., J. Lawson, and A. Radziwill (2010). Current account imbalances in the euro area: A comparative perspective. OECD Economics Department Working Papers (826). Barnett, A. and R. Straub (2008). What drives US fluctuations? ECB Working Paper (959). Barrios, S., P. Iversen, M. Lewandowska, and R. Setzer (2009). Determinants of intra-eurozone bond spreads during the financial crisis. European Economy Economic Papers (388). Belke, A. and C. Dreger (2011). Current account imbalances in the euro area: Catching up or competitiveness? DIW Berlin Discussion Papers (1106). Bems, R., L. Dedola, and F. Smets (2007). Us imbalances: The role of technology and policy. Journal of International Money and finance 26 (4), Benkwitz, A., H. Lütkepohl, and J. Wolters (2001). Comparison of bootstrap confidence intervals for impulse responses of german monetary systems. Macroeconomic Dynamics 5 (01), Binning, A. (2013). Underidentified svar models: A framework for combining short and long-run restrictions with sign-restrictions. 17

19 Blanchard, O. and F. Giavazzi (2002). Current account deficits in the euro area: the end of the feldstein-horioka puzzle? Brookings papers on economic activity, Blanchard, O. J. and D. Quah (1989). The dynamic effects of aggregate demand and supply disturbances. American Economic Review 79 (4). Candelon, B. and H. Lütkepohl (2001). On the reliability of chow-type tests for parameter constancy in multivariate dynamic models. Economics letters 73 (2), Canova, F. and G. D. Nicoló (2002). Monetary disturbances matter for business fluctuations in the g-7. Journal of Monetary Economics 49 (6), Ca Zorzi, M. and M. Rubaszek (2008). On the empirical evidence of the intertemporal current account model for the euro area countries. ECB Working Paper Series (895). Fry, R. and A. Pagan (2011). Sign restrictions in structural vector autoregressions: a critical review. Journal of Economic Literature 49 (4), Golub, G. H. and C. F. Van Loan (2012). Matrix computations, Volume 3. JHU Press. Gros, D. (2011). External versus domestic debt in the euro crisis. CEPS Policy Brief (243). Holinski, N., C. Kool, and J. Muysken (2012). Origins of persistent macroeconomic imbalances in the euro area. Federal Reserve Bank of St. Louis Review (94). IMF (2010, April). Meeting new challenges to stability and building a safer system. Global FInancial Stability Report. Jaumotte, F. and P. Sodsriwiboon (2010). Current account imbalances in the southern euro area. IMF Working Papers 139 (201), Lane, P. R. and B. Pels (2012). Current account imbalances in europe. Lütkepohl, H. (2011). Vector autoregressive models. EUI Department of Economics Working Paper (30). Mongelli, F. P. and C. Wyplosz (2008). The euro at ten: unfulfilled threats and unexpected challenges. pp

20 Obstfeld, M. and K. Rogoff (1995). The intertemporal approach to the. Handbook of international economics 3, Rubio-Ramirez, J. F., D. F. Waggoner, and T. Zha (2010). Structural vector autoregressions: Theory of identification and algorithms for inference. The Review of Economic Studies 77 (2), Schnabl, G. and T. Wollmershäuser (2013). Fiscal divergence and imbalances in europe. Cesifo Working Paper (4108). Uhlig, H. (2005). What are the effects of monetary policy on output? Results from an agnostic identification procedure. Journal of Monetary Economics 52 (2), Walters, A. (1990). Sterling in danger: the economic consequences of pegged exchange rates. Fontana. Wyplosz, C. (2010). Ten years of emu: successes and puzzles. Spain and the euro, Wyplosz, C. (2013). Eurozone crisis: It s about demand, not competitiveness. 19

21 7 Appendix Figure 4: Data used in the VAR models for Spain, Italy, Portugal, Ireland and Germany. Spain gross domestic product Italy gross domestic product Portugal gross domestic product Ireland gross domestic product Germany gross domestic product Figure 5: Impulse responses for the Italian model, 68 % confidence bands (blue=pre-emu regime, red=emu regime)

22 Figure 6: Impulse responses for the Portuguese model, 68 % confidence bands (blue=pre-emu regime, red=emu regime) Figure 7: Impulse responses for the Irish model, 68 % confidence bands (blue=pre-emu regime, red=emu regime)

23 Figure 8: Impulse responses for the German model, 68 % confidence bands (blue=pre-emu regime, red=emu regime)