The euro s effect on trade balance dynamics

Transcription

1 The euro s effect on trade balance dynamics Mika Nieminen Juha Junttila September 27, 2015 Abstract During the pre-emu period changes in real effective exchange rate or faster-than-tradingpartners growth rates Granger caused changes in trade balance in most of the EMU-12 countries. However, our data driven article provides evidence that after the adoption of euro, these Granger causalities disappeared. We decompose trade balances into intra balances (trade balance vis-à-vis the euro area) and extra balances (trade balance vis-à-vis the rest of the world), and find that the disappearance of dynamic feedback effect typically occurred in the intra balances rather than in the extra balances. Our results imply that debtor countries cannot reduce their trade deficits in the short-run by enhancing price competitiveness or by adopting austerity policies. Only in Finland the trade balance has improved (or deteriorated) as a result of real effective exchange rate depreciation (or appreciation) during the EMU period. Keywords: external adjustment; J-curve; competitiveness; European monetary union JEL classification: F32; F45 The authors would like to thank the participants of the 47th Money, Macro and Finance Research Group Annual Conference at Cardiff University for their comments and Zsolt Darvas (Bruegel) for providing the real effective exchange rate data with specific weighting matrices. Mika Nieminen is grateful for financial support provided by the Björn Savén Finnish American Scholarship and the OP-Pohjola Group Research Foundation. This study is part of the research agenda of the JSBE Research Group on International Macro and Finance (JyIMaF). Jyväskylä University School of Business and Economics, University of Jyväskylä, P.O. 35, FI-40014, Finland Corresponding author: mika.p.nieminen@jyu.fi 1

2 1 Introduction When the euro was introduced, widening current account imbalances were not seen as a problem, but rather as a natural consequence of the economic integration (see, e.g., Blanchard and Giavazzi 2002). After a decade much more cautious view has been adopted in the European Union. A legislative package and a surveillance procedure for the prevention and correction of macroeconomic imbalances were enforced in The Macroeconomic Imbalance Procedure (MIP) stresses the importance of external balance and competitiveness. 2 Many of the factors in the MIP can be viewed to be based on the standard Mundell-Fleming model, according to which external adjustment can occur through expenditure switching (a change in exchange rate) or expenditure shifting (a change in domestic demand). European Commission (2010a, 8) claims that differences in the domestic demand and price competitiveness have contributed to the divergence of current account balances, and intra-euro area imbalances have been a large part of this divergence (see, e.g., ECB 2013, 69, or Darvas 2012). We use standard VAR methodology and try to find answers to the three following research questions: First, have the changes in real effective exchange rates or growth differentials (i.e., in the difference between domestic and foreign GDP per capita growth rates) Granger caused changes in the trade balances in EMU-12 countries. Second, did the dynamic feedback change, when the countries adopted the euro. Third, if there indeed was a change, did it occur in the intra balance or in the extra balance. 3 The introduction of euro potentially changed the co-evolution of real effective exchange rates and intra-euro area trade balances (see Figure A1). To our knowledge no article has addressed all of these three questions together. 1 Regulation (EU) No 1176/2011 of the European Parliament and of the Council of 16 November 2011 on the prevention and correction of macroeconomic imbalances (OJ L 306, , pp ) and Regulation (EU) No 1174/2011 of the European Parliament and of the Council of 16 November 2011 on enforcement measures to correct excessive macroeconomic imbalances in the euro area (OJ L 306, , pp. 8-11). 2 The scoreboard consists of current account balance, net international investment position, export market shares, nominal unit labor costs, real effective exchange rates, private sector debt, private sector credit flow, changes in the house price index, general government sector debt, and unemployment rate and there is a threshold value for each of these indicators. These indicators are claimed to focus on the most relevant dimensions of macroeconomic imbalances and competitiveness losses. (European Commission 2012a). 3 Intra balance is the trade vis-à-vis the EMU-12 countries and extra balance is the trade balance vis-à-vis the rest of the world. 1

3 Berger and Nitsch (2014) as well as Zemanek et al. (2010) are related to our paper. Berger and Nitsch (2014) use the same identification scheme as in the so-called Rose literature (i.e., the effect of common currency on trade volumes). 4 They find that after the introduction of the euro the intra-euro area imbalances became both larger and more persistent. Bilateral surpluses seem to have been decreasing both in relation to real exchange rate appreciations and growth differentials. However, there are some fundamental differences between Berger and Nitsch (2014) and our paper. First, they have annual data whereas we use quarterly data. More importantly, they do not aggregate bilateral trade balances into the intra and extra balances. Furthermore, they estimate the model for a panel data, whereas we examine one country at a time. In addition, they do not discuss the potential problems caused by nonstationarity even though their time series are very long. Zemanek et al. (2010) employ a dynamic panel model (system GMM estimator) to find out whether structural reforms or private sector adjustments affect bilateral trade balances. They find for example that the changes in unit labor costs are statistically insignificant in explaining the changes in bilateral trade balances. Our paper complements these two articles and offers a bit more detailed results and discussion for some individual EMU countries regarding the effects of the single common currency on their intra and extra trade balances. According to undergraduate textbooks the response of country s trade balance to the real exchange rate depreciation follows a J-curve. Initially net exports deteriorate, but once the import and export volumes have time to adjust, the overall effect is positive. However, there is no strong empirical evidence for the J-curve effect (see, e.g., Bahmani-Oskooee and Ratha 2004). For example, in the above mentioned MIP scoreboard a -/+5% change over three years is the threshold value for the real effective exchange rate. The threshold is defined for a change over a three-year period and therefore a cross-section estimation with multi-year averages would be one possible framework for our analysis. Indeed some studies have run cross-section regressions 4 Baldwin (2006) is a thorough review of the Rose literature. 2

4 in which changes in the trade balance or current account balance are explained by changes in the real effective exchange rate and GDP per capita growth rate (see, e.g., Estrada et al. 2013, Table 1). However, these regressions have some problems. First, it is very difficult to claim that real exchange rate could be considered as exogenous to trade balances. Secondly, it is very likely that the countries differ with respect to adjustment dynamics. When we ran such cross-section regressions with EU-15 countries using 5-year rolling averages, we observed one additional problem: the results are very sensitive to the time interval (see Figure A2). By using conventional VAR analysis for the pre-emu and EMU periods we can avoid these problems. The remainder of this paper is organized as follows: In Section 2 we will briefly go through the empirical literature in which the relation between the trade balance, the real exchange rate and foreign demand has been studied. Our emphasis is on the discussion about the factors that have contributed to diverging current account balances in the euro area. In Section 3 we describe our data and present our empirical methodology. In Section 4 we give our main results. We will take a closer look on Finland, Greece, Portugal and France as these are the countries for which we can detect a clear change in the trade balance dynamics and for which we can pin down the change with our decomposition. We will also discuss Germany and Spain briefly because these two are some sort of anomalies. Section 5 concludes. 2 Literature Bahmani-Oskooee and Ratha (2004) summarize the early literature on the J-curve phenomenon. Their overall assessment is that the short-run response of trade balance to devaluation is country specific and does not follow any specific pattern. In the long-run the effect is positive and this result appears more often with bilateral data than aggregate data. (Bahmani-Oskooee and Ratha 2004.) It is important to realize that the previous J-curve studies differ in many ways: for the set of countries, the time period, in terms of the utilized empirical methodology, using bilateral or aggregate trade data, and using exchange rate or the terms of trade. Rose and 3

5 Yellen (1989) examine the hypothesis of J-curve with US data. When using bilateral data, they cannot find any effect from exchange rate to trade balance. With aggregate data their results depend on the estimation method. According to Rose and Yellen previous studies had failed to take into account the potential simultaneity of trade balance, exchange rate and output, as well as the nonstationarity of the time series. They conclude that the hypothesis that there is no response of the trade balance to the real exchange rate, in either the short run or long run, is just as consistent with the data as the J-curve hypothesis. Demirden and Pastine (1995) stress the importance of feedback effects between the trade balance, real exchange rate, domestic income and foreign income in a flexible exchange rate regime. Hence, they assert that VAR methodology in which all variables are allowed to be endogenous is well suited for this purpose. More recently some studies have utilized structural VAR models for exchange rates and current account balances that are based on the implications of the intertemporal approach to current account. Lee and Chinn (2006) estimate a VAR model involving exchange rates and current account balances. Their key assumption is that temporary effects (nominal shocks) have no long-run effects on the real exchange rate. 5 However, Fisher and Huh (2002) point out that in the more recent intertemporal models with sticky prices nominal shocks can have a long-run effect on the real exchange rate as well as on the trade balance. Gourinchas and Rey (2014, 586) remark that actually the key predictions of the intertemporal approach to the current account have been rejected several times by the data. 6 European Central Bank (2012) is an extensive analysis on the factors that contributed to the diverging current account balances in the euro area. Some EMU member countries suffered competitiveness losses (measured by the unit labor cost (ULC) based real effective exchange rate). The appreciation of ULC-based real effective exchange rate indicates that the development of labor costs was not driven by changes in productivity. A wage determination 5 Giuliodori (2004) extends their model by including a demand shock. 6 Nason and Rogers (2006) is a good summary of the potential reasons for the shortcomings. See also Bergin (2006), Kano (2008), and Campa and Gavilan (2011). 4

6 mechanism with wage spillovers from non-traded or public sector to the traded-sector could explain this disconnection. Given the high taxes on labor incomes, a country can improve its competitiveness by fiscal devaluation, i.e., shifting from direct taxes and social security contributions towards indirect taxes. On the other hand, productivity growth can increase competitiveness via improving price competitiveness (better process efficiency, improved skills, etc.) or better non-price competitiveness (higher product quality etc.). The only problem is that it is difficult to measure the ability to innovate. 7 Based on disaggregated sectoral data, the contributions of price and non-price competitiveness factors on trade balances seemed to have been equally important for most of the EMU member countries. Model simulations on four different models ranging from a Dynamic Stochastic General Equilibrium (DSGE) model to an empirical Global Vector Autoregressive Model (GVAR) suggest that a 5% to 10% temporary reduction in relative wages is needed for the current account balance to improve by 1% of GDP in the medium term. However, the short-run impulse responses vary substantially across models. The peak of the current account response could be 3 quarters or 13 quarters after the shock depending on the model. (European Central Bank 2012.) Collignon and Esposito (2014) point out that wages do not measure the total cost structure of economy. Hence, they develop a competitiveness index which includes also capital efficiency and profit rates. However, there is no large difference between the real effective exchange rate and their own index in the explanatory power for the trade balances. According to European Commission (2010b) foreign demand and real effective exchange rate accounted only for a half of the variation in exports of the EMU member countries during Furthermore, the correlation between exports and real exchange rate has actually been positive. This has been interpreted as an evidence for the fact that price competitiveness is only one of the factors determining the export performance. (European Commission 2010b, 7 See Nieminen (2015) in which he found that there is a strong positive link between Hofstede s Individualism index and intra-euro area trade balances. Based on Gorodnichenko and Roland (2010) the Individualism index is a very good proxy for the ability to innovate. 5

7 29.) If the relation between real effective exchange rate and exports is not clear this relation might be even less clear with net exports, because the import content of export goods can be very high (see European Commission 2012b, 31, Graph 3.1, or European Central Bank 2012, 30-32). Wyplosz (2013) actually claims that the competitiveness narrative of the Eurozone crisis is misleading and based on the faulty use of data. He has three arguments for this interpretation: First, EMU member countries do not only compete with each other and therefore unit labor cost should be measured with respect to all countries not just relative to the EMU countries. Second, changes in total economy unit labor costs might reflect the changes in nontraded goods sector. Third, there is no reason to set indices to 100 at some arbitrary year as if real effective exchange rates would have been in equilibrium at this arbitrary year. It is more reasonable to assume that the real effective exchange rates are in equilibrium in the long-run. If these points were taken into consideration, the divergence in competitiveness would probably be much less dramatic. On the other hand, the competitiveness narrative does not explain why inflation has been higher in Southern European countries. It is very likely that changes in competitiveness have been endogenous and driven by domestic demand shocks. Countries that entered the EMU with above-average inflation rates had lower-than-average real interest rates and this resulted in credit booms and high domestic demand. (Wyplosz 2013.) Related to this view, Fratzscher et al. (2010) build a Bayesian structural VAR model and show that asset market reactions (equity market shocks and housing price shocks) have been much more important than the behavior of exchange rate in explaining the US current account balance. 3 Data and methodology We use quarterly data on trade balances, real exchange rates and GDP per capita growth rates. We build three different three-variable VARs for each EMU-12 countries excluding Luxembourg 6

8 and Belgium. 8 The length of the sample period varies from 63 quarters (1998Q1-2013Q3, Ireland) to 171 quarters (1971Q1-2013Q3, France, Germany and Spain). For Ireland and Austria we cannot compare the pre-emu period to the EMU period, because for these two countries the pre-emu sample length is so short for them, 4 quarters for Ireland and 40 quarters for Austria. Most of our time series seem to behave like nonstationary processes, even though there seems to be some discrepancy on whether the real effective exchange rates are stationary or not (see Table A1). There is strong seasonality not just in the output series but also in the trade balance series. In order to make the series stationary and to get rid of the seasonality we use year-to-year changes. These series are stationary. 9 We use the CPI-based real exchange rates, because this is the indicator that was included to the Macroeconomic Imbalance Procedure scoreboard. 10 Another option would have been to use the ULC-based real exchange rates. However, these series are typically rather short. In the first set of VAR models we used the changes in aggregate trade balance (as ratios to GDPs), the changes in CPI-based real effective exchange rate, and the differences between the domestic GDP per capita growth rate and the GDP per capita growth rate in the world economy. 11 Hence, the set of partner countries consists of the whole world. In the second set we use changes in the intra balance (as ratios to GDPs), changes in the CPI-based real exchange rate against the EMU-12 countries, and the differences between the domestic GDP per capita growth rate and the GDP per capita growth rate in the rest of the EMU-12 countries. In these models the set of partner countries consists of the EMU-12 countries. In the third set of VAR models we use the changes in extra balance (as ratios to GDPs), changes in the CPI-based real exchange rate against the non-emu-12 countries, and the differences between the domestic GDP per capita growth rate and the GDP per capita growth rate in the non-emu-12 countries. 8 The problem is that prior to year 1997 there is no data for these two countries separately in the IMS s Direction of Trade Statistics. However, both the Luxembourg and the Belgium are included as partner countries, when intra balances or real exchange rates against the EMU-12 countries are calculated. 9 For Ireland the series are so short that the null hypothesis of nonstationarity is not always rejected. 10 Note that in order to calculate the real exchange rates against an arbitrary group of countries (EMU-12 countries or non-emu-12 countries), one needs to re-scale the trade weighting matrix. 11 See details from Table A2. 7

9 In this case the set of partner countries consists of the non-emu-12 countries. The lag structure is always selected using the Bayesian Information Criterion of Schwarz. Consequently, in all stages we estimate the following VAR(p) model: x t = ν + A 1 x t A p x t p + u t, (1) where x t = ( reer t, y t y t, tradebalance t ) is a column vector, the A i are coefficient matrices, ν is a column vector of intercept terms, and u t is an innovation process. The fundamental difference between the three different sets of VAR models is that in each case the set of partner countries is different (the whole world, the EMU-12 countries or the rest of the world) and this has an effect to all included variables not just the trade balance. We use conventional VAR modelling, because it would be difficult to derive identifying restrictions for the effects of intra and extra balances. Lee and Chinn (2006) focused on current account balances and real effective exchange rates and they assumed that the temporary effects (nominal shocks) have no long-run effects on the real exchange rates. Fisher and Huh (2002) questioned such identification restrictions. Actually, it could be difficult to explore the effects of nominal shocks on trade balances using our sample, because all EMU countries face the same monetary policy. Furthermore, Lee and Chinn (2006) claimed that the ordering of variables in the VAR models for these type of analyses is arbitrary in the Choleski factorization. We use the following ordering: changes in the real effective exchange rate, growth differentials, and changes in trade balance. This implies that the real exchange rate shock has a contemporaneous effect on the trade balance, but not vice versa. We tested also an alternative ordering: growth differentials, changes in the trade balance, and changes in the real effective exchange rate. Concerning the aggregate trade balances none of our results changed. This robustness is based on the fact that the correlations between the different shocks are typically very low. 8

10 4 Results 4.1 Testing Granger causality before and after the adoption of euro Our VAR analysis indicates that during the pre-emu period the trade balances did in most cases respond to changes in the real effective exchange rate or growth differentials (see Table 1). This holds at the aggregate level for all countries except Germany and Italy. 12 As could be anticipated real effective exchange rate appreciation and faster-than-trading-partners economic growth have had a negative effect on the trade balance. However, during the EMU period we can find such dynamic feedback effects only for Finland. The decomposition of trade balance into intra and extra balances indicates that the disappearance of the Granger causality effects in aggregate trade balances typically resulted from intra balances. 13 With respect to the real effective exchange rate this holds for Portugal and France 14 and with respect to the growth differential for Finland. Greece is an exception, because it is the extra balance that does not react to growth differentials anymore. However, for Greece the EMU period is shorter and therefore it is possible that the disappearance of Granger causality is partly due to fewer observations. 15 The dynamics changed also in Netherlands and Spain, but our decomposition fails to pin down this change. 16 The Finnish trade balance has responded to changes in the real effective exchange rate during the EMU period. 17 Decomposition of the trade balance reveals that this change took place in the extra balance dynamics. Typically the lagged values of trade balances have a strong positive effect on trade balances. This implies that external adjustment tends to be slow. 18 The 12 This holds also for Germany and Italy, if we take into consideration intra and extra balances separately. 13 For Germany there is a clear change in extra balances, but this does not show up in the aggregate trade balance. 14 The change in intra balance is larger than in extra balance. According to VAR(1) model the real effective exchange rate appreciation Granger caused extra balance deterioration at the 10% significance level in the pre- EMU period. Hence the effect was much stronger on the intra balance. 15 Because Greece s EMU period is shorter we report the whole sample results for Greece in Table Netherlands has had a huge intra surplus (18.6% of GDP on average ) and a huge extra deficit (12.4% of GDP) during the EMU period. This is probably due to the Rotterdam effect and this can mix up these results. 17 The Bayesian Information Criterion suggests VAR(2) model for Finland during the EMU period. However, exactly equally strong Granger causality comes up from VAR(1) model. Therefore the change between the pre-emu period and the EMU period does not result from the different lag structure. 18 The speed of adjustment for trade balance shocks is presented in Figure A3. 9

11 results on Granger causality for the whole sample period are presented in Table A3. It is also interesting to see whether changes in trade balances have Granger caused changes in the real effective exchange rates or growth differentials (see Table A4.) Table 1 Consequently, we can find some general patterns, but there are also considerable differences between the analyzed countries. The relative importance of intra and extra balances is naturally related to the share of intra trade which varies considerably across the EMU-12 countries (see Table A5). For Finland and Greece the share of intra trade is low. This probably explains at least to some extent why for example for Greece the changes in Granger causality regarding extra balances have resulted in same changes in Granger causality for aggregate trade balances. However, this does not tell anything about the reason why the dynamics changed in the first place. One interesting finding was that for some countries the intra balance Granger causes extra balance (or vice versa). For example in Spain there has been a strong positive feedback from the intra balance to the extra balance during the EMU period. This implies that if the intra surplus (deficit) has been large in the current period, the extra surplus (deficit) tends to be large in the next period. In other words, there are some factors that both the real effective exchange rate and the GDP per capita growth rate fail to measure which are important for the extra balance and which the intra balance is able to capture. Interestingly this did not hold to opposite direction (from extra to intra) and not during the pre-emu period. The same dynamics but with an opposite sign was found for the Netherlands during the EMU period. This is very likely an outcome of transit trade: if Netherlands is importing (exporting) a lot in the current period from (to) EU-12 countries, it will be exporting (importing) a lot to (from) the rest of the world in the next period. Again this does not hold to opposite direction. For Germany there was a strong positive feedback effect from extra balance to intra balance in the 10

12 EMU period. For the rest of the countries such effects were not found. Next we will take a closer look on Finland, Greece, Portugal and France as these are the countries for which we can detect a clear change in the trade balance dynamics and for which we can pin down the change with our decomposition. We will also briefly discuss Germany and Spain because according to the results reported in Table A4 positive intra balance shocks Granger cause devaluation of the real effective exchange rate in these two countries. 4.2 Finland In Finland the dynamics of the trade balance apparently changed, when the country adopted the euro currency. Impulse responses for the aggregate trade balance are shown in Figure 1. The red dotted lines represent the 95% error bands (based on a Monte Carlo integration with draws). Faster-than-trading-partners economic growth has had a negative effect on the trade balance during the pre-emu period, but this has not been the case during the EMU period (see also Table 1). This change occurred clearly in the intra balances (see Figure 2, rows 1 and 2). The dynamic response of aggregate trade balance to the real effective exchange rate also changed. During the EMU period the real effective exchange rate has had a deteriorating effect on the trade balance and this change took place specifically in the extra balances (see Figure 2, rows 3 and 4). This implies that for Finland the exchange rate of euro against other currencies is an important determinant of external adjustment. To be exact, Figure 1 indicates that a 5% appreciation in the CPI-based real effective exchange rate has a 1.5% of GDP negative effect on the level of the Finnish trade balance. In the European Central Bank (2012, 51) the authors conclude that an improvement in the current account balance of 1% of GDP in the medium term requires a temporary reduction in wages relative to competitors of around 5% to 10%. Therefore, if we assume that changes in wages represent a one-to-one changes in the real effective exchange rate, we can conclude that our result is consistent with the European Central Bank (2012). Furthermore, one should also notice that during the EMU period changes in the 11

13 trade balance Granger caused changes in the real effective exchange rate (see Figure A4). The current trade surplus (deficit) has a positive (negative) effect on future real effective exchange rate, so also the sign of the effect is theoretically reasonable. Figure 1 Figure Greece Greece ran huge trade deficits throughout the whole sample period (1980Q1-2013Q3): at the level of 13.1% of GDP on average. For Greece the experience of euro is shorter and therefore it makes sense to take a look at the whole sample period also. Growth differential seems to have had an effect on the Greek trade balance, whereas the real effective exchange rate seems not (see Table 1 and Figure 3). Consequently Greece can improve its trade balance via slowerthan-trading-partners economic growth, but not via enhancing its price competitiveness. 19 This dynamic effect on the aggregate trade balances can be pinned down to the extra balance, even though it seems that the dynamic feedback from growth differential has become weaker during the EMU period. On the other hand, it is difficult to say if anything has really changed, because for Greece there are fewer observations from the EMU period and therefore the error bands are inevitable wider. It is striking to see how much slower the external adjustment for trade balance shocks has become during the EMU period for Greece. This is noticeable both in the extra and intra balances. This could reflect the fact that it became easier for Greece to finance its deficits. After Greece had adopted the euro, the foreign investors did not have to consider currency risk. On the other hand, Target balances have made the external adjustment of deficit countries smoother 19 Actually we tested also using the ULC-based real effective exchange rate for the EMU period, but the result did not change. This robustness check was made for the EMU period only, because the ULC-based real effective exchange rates time series were shorter than the CPI-based series. 12

14 (see, e.g., Cour-Thimann 2013, 23). 20 When the Eurozone crisis broke out the private capital flowed out of the debtor countries, but the widening of the Target imbalances compensated this. This could explain why the adjustment in intra balance became slower. Figure Portugal and France In Portugal and France the changes in real effective exchange rate Granger caused the changes in trade balances during the pre-emu period but not during the EMU period (see Table 1 and Figure A5). For both of these countries the change took place in the intra balance. When looking at the impulse response functions, this change is evident for Portugal (Figure A5, rows 1-4). For France it is obvious that the intra balance dynamics changed (Figure A5 rows 7 and 8), whereas this change cannot be seen so clearly from the impulse response functions with aggregate data (Figure A5 rows 5 and 6). In France there was weak positive Granger causality from the trade balance to the real effective exchange rate during the pre-emu period (see Table A4). 4.5 Germany and Spain Germany and Spain are interesting anomalies because during the EMU period a positive (negative) intra balance shock has Granger caused a devaluation (revaluation) in the real effective exchange rate in these countries (see Table A4 and Figure A6). Such a dynamic feedback tends to destabilize trade imbalances further from the balance. Usually this relation is positive (see for example Figure A4 for Finland in the EMU period). This could be one explanation for Germany s large and persistent trade surplus and Spain s large and persistent trade deficits. 20 At the end of the year 2012, Greece s Target balance deficit was 50.6% of its GDP which was the highest debt ratio among EMU-12 countries. Data on the Target balances can be found from CESifo < http : // group.de/dms/ifodoc/docs/politikdebatte/c Haftungspegel/T arget countries/t arget countries xlsx > and data on GDPs from Eurostat < http : //ec.europa.eu/eurostat/data/database >. 13

15 5 Conclusions Using a standard VAR model and quarterly data for the EMU-12 countries we analyze the relationship between trade balances, real effective exchange rates and GDP per capita growth differentials during the pre-emu and EMU periods. Contrary to standard undergraduate textbook presentations we cannot find robust evidence that the trade balances would be decreasing in the real effective exchange rate and growth differentials. However, this result is actually consistent with some previous studies. Our main contribution was to show that the introduction of common currency affected the trade balance dynamics and to pin down this change by decomposing the aggregate trade balances into the intra (trade balances vis-à-vis the euro area) and extra balances (trade balances vis-à-vis the rest of the world). In most of the cases the observed change has resulted from the intra-euro area trade balances. We analyzed Finland, Greece, Portugal and France more closely, because our identification scheme was perhaps the most successful for these countries. In addition, we found out that both in Germany and Spain there is destabilizing dynamics as positive (negative) intra balance shocks devaluate (revaluate) real effective exchange rate. In Finland, faster-than-trading-partners GDP per capita growth rate has had a negative effect on the trade balance during the pre-emu period, but not during the EMU period anymore, and this change has occurred specifically in the intra-emu balances. In addition, during the EMU period the real effective exchange rate has had a deteriorating effect on the trade balance and this comes through the dominating effect of the extra balance. This is plausible, because Finland has a high foreign trade activity with non-emu countries. We obtain empirical evidence that for the Finnish economy, in terms of international competitiveness, the real exchange rate of euro against other currencies is important for the external adjustment also nowadays. For Greece the relative growth rate seems to have had a negative effect on the trade balance, but it remains a bit unclear whether this effect has become weaker. We cannot find any evidence that 14

16 the price competitiveness in terms of the CPI-based real effective exchange rate would have had an effect on Greek external balance. This is an empirical finding that requires more thorough analyses in the future, not the least due to the enormous amount of political and economic debates that surround the Greek survivorship story at the moment. In media internal devaluation and austerity policies are usually the only suggested remedies for the external adjustment. This same belief is reflected in the European Commission s Macroeconomic Imbalances Procedure (MIP) which stresses the importance of price competitiveness. Our unconventional results call for a closer investigation of the forces behind the changes in the connections between trade balances, the real exchange rate and growth differentials. In addition to measuring the relevant variables perhaps in a different way, for example using the real unit labor costs based real effective exchange rate, the time series properties of the analyzed variables should probably be scrutinized more carefully. All the variables analyzed here seemed to behave like unit root processes in levels, so instead of focusing on year-to-year percentage changes (that are clearly interpretable in terms of growth rates), one should perhaps use this set of variables in levels in time series analysis, and attempt to separate their long-run and short-run economic connections e.g. based on cointegration and error correction techniques. Furthermore, it is obvious that the explanatory power of the analyzed fundamental factors (real exchange rate and GDP per capita growth) could be examined also by taking into account their possible endogeneity via other than the utilized VAR estimation methods, like the GMM estimation, and hence, by instrumenting them on a set of relevant instrumental variables. 6 Appendix: Additional tables and figures Table A1 Table A2 Table A3 15

17 Table A4 Figure A1 Figure A2 Figure A3 Figure A4 Figure A5 Figure A6 References [1] Bahmani-Oskooee, M., Ratha, A The J-curve: a literature review. Applied Economics 36 (13), [2] Baldwin, R The euro s trade effects. European Central Bank Working Paper Series No 594. [3] Berger, H., Nitsch, V Wearing corset, losing shape: the euro s effect on trade balances. Journal of Policy Modeling 36 (1), [4] Bergin, P. R How well can the New Open Economy Macroeconomics explain the exchange rate and current account?. Journal of International Money and Finance 25 (5), [5] Blanchard, O., Giavazzi, F Current account deficits in the euro area: the end of the Feldstein-Horioka puzzle? Brookings Papers on Economic Activity 2002:2, [6] Campa, J. M., Gavilan, A Current accounts in the euro area: an intertemporal approach. Journal of International Money and Finance 30 (1), [7] Collignon, S., Esposito, P Unit labour costs and capital efficiency in the euro area - a new competitiveness indicator. In: Collignon, S., Esposito, P. (Eds.), Competitiveness in the European economy. Routledge, pp [8] Cour-Thimann, P Target balances and the crisis in the euro area. CESifo Forum 14, Special Issue. [9] Darvas, Z Intra-euro rebalancing is inevitable, but insufficient. Bruegel Policy Contribution, Issue 2012/15. [10] Demirden, T., Pastine, I Flexible exchange rates and the j-curve: an alternative approach. Economics Letters 48 (3-4), [11] Estrada, A., Gali, J., Lopez-Salido, D Patterns of convergence and divergence in the euro area. IMF Economic Review 61 (4),

18 [12] European Central Bank Competitiveness and external imbalances within the euro area. ECB Occasional Paper No 139. [13] European Central Bank Intra-euro area trade linkages and external adjustment. ECB Monthly Bulletin 01/2013. [14] European Commission. 2010a. The impact of the global crisis on competitiveness and current account divergences in the euro area. Quarterly Report on the Euro Area 9 (1). [15] European Commission. 2010b. Special topics in the euro-area economy: Assessing the sources of non-price competitiveness. Quarterly Report on the Euro Area 9 (2), [16] European Commission. 2012a. Scoreboard for the surveillance of macroeconomic imbalances. European Economy Occasional Papers 92. [17] European Commission. 2012b. A closer look at some drivers of trade performance at Member State level: Import content of exports. Quarterly Report on the Euro Area 11 (2), [18] Fisher, L. A., Huh, H.-S Real exchange rates, trade balances and nominal shocks: evidence for the G-7. Journal of International Money and Finance 21 (4), [19] Fratzscher, M., Juvenal, L., Sarno, L Asset prices, exchange rates and the current account. European Economic Review 54 (5), [20] Giuliodori, M Nominal shocks and the current account: a structural VAR analysis of 14 OECD countries. Review of World Economics 140, (4), [21] Gorodnichenko, Y., Roland, G Culture, institutions and the wealth of nations. NBER Working Paper Series. [22] Gourinchas, P.-O., Rey, H External adjustment, global imbalances, valuation effects. In: Gopinath, G., Helpman, E., Rogoff, K. (Eds.), Handbook of International Economics, Volume 4. North-Holland, pp [23] Kano, T A structural VAR approach to the intertemporal model of the current account. Journal of International Money and Finance 27 (5), [24] Lee, J., Chinn, M. D Current account and real exchange rate dynamics in the G7 countries. Journal of International Money and Finance 25 (2), [25] Nason, J. M., Rogers, J. H The present-value model of the current account has been rejected: round up the usual suspects. Journal of International Economics 68 (1), [26] Nieminen, M Trade imbalances within the euro area and with respect to the rest of the world. Economic Modelling 48, [27] Rose, A. K., Yellen, J. L Is there a J-curve? Journal of Monetary Economics 24 (1), [28] Wyplosz C The eurozone crisis and the competitiveness legend. Asian Economic Papers, 12 (3), [29] Zemanek, H., Belke, A., Schnabl, G Current account balances and structural adjustment in the euro area. International Economics and Economic Policy 7 (1),

19 Table 1. Granger causality test of the VAR models (p-values of rejecting the null hypotheses) Set of partner countries: The whole world EMU-12 countries The rest of the world Time period: pre-emu EMU pre-emu EMU pre-emu EMU Null hypothesis (below): reer t-1 tradebalance t neg pos neg FI y t-1 - y* t-1 tradebalance t 9 neg neg neg tradebalance t-1 tradebalance t pos pos pos FR DE GR reer t-1 tradebalance t neg neg mix y t-1 - y* t-1 tradebalance t neg mix pos tradebalance t-1 tradebalance t pos pos pos pos mix pos reer t-1 tradebalance t neg y t-1 - y* t-1 tradebalance t neg tradebalance t-1 tradebalance t pos pos pos pos pos pos reer t-1 tradebalance t y t-1 - y* t-1 tradebalance t neg neg tradebalance t-1 tradebalance t pos pos pos 2 neg pos GR 1981Q1-2013Q3 reer t-1 y t-1 - y* t-1 tradebalance t-1 tradebalance t tradebalance t tradebalance t neg pos neg 1 neg IT NL PT ES reer t-1 tradebalance t neg y t-1 - y* t-1 tradebalance t tradebalance t-1 tradebalance t pos pos pos pos pos pos reer t-1 tradebalance t y t-1 - y* t-1 tradebalance t neg tradebalance t-1 tradebalance t pos pos pos pos pos reer t-1 tradebalance t neg neg y t-1 - y* t-1 tradebalance t mix tradebalance t-1 tradebalance t pos pos pos pos mix pos reer t-1 tradebalance t 3 neg y t-1 - y* t-1 tradebalance t tradebalance t-1 tradebalance t pos pos pos pos pos pos Notes: The null hypothesis is such that real effective exchange rate (or relative GDP per capita or lagged value of trade balance) does not Granger cause trade balance. We report the p-value of rejecting the null hypothesis. Bold font indicates that the null hypothesis is rejected at the 10% significance level. The superscript "pos" indicates a positive Granger causality. The superscript "neg" indicates a negative Granger causality. The superscript "mix" indicates that the signs of the parameter values in higher order models alternate. All models include an intercept term. All variables are measured as yearto-year changes. Lag-lengths in VAR-models are selected based on Bayesian information criterion. For the most part all models are VAR(1)-models. The only exceptions are the trade balance model for Finland during the EMU period (VAR(2)), the extra balance model for France during the pre-emu period (VAR(2)), and the extra balance model for Portugal during the pre-emu period (VAR(4)). For the higher-order VAR models the joint significance of lagged values is reported. Abbreviations: Finland (FI), FR (France), Germany (DE), Greece (GR), Italy (IT), Netherlands (NL), Portugal (PT), Spain (ES), year-to-year changes in the real effective exchange rate ( reer), year-to-year changes in the relative GDP-per-capita growth rate ( y- y*). 18

20 Figure 1: Impulse response functions for Finland (aggregate trade balance) Response of tradebalance with 95% error bands, Finland, 1976Q1-1998Q4 0 0 reer tradebalance y- y* tradebalance Response of tradebalance with 95% error bands, Finland, 1999Q1-2013Q3 reer tradebalance y- y* tradebalance Cumulative response of tradebalance with 95% error bands, Finland, 1976Q1-1998Q y- y* tradebalance reer tradebalance Cumulative response of tradebalance with 95% error bands, Finland, 1999Q1-2013Q3 reer tradebalance y- y* tradebalance

21 Figure 2: Impulse response functions for Finland (intra and extra balances) 0 75 Response of intrabalance with 95% error bands, Finland, 1976Q1-1998Q4 0 0 reer intrabalance 75 y- y* intrabalance 75 intrabalance intrabalance Response of intrabalance with 95% error bands, Finland, 1999Q1-2013Q3 reer intrabalance y- y* intrabalance intrabalance intrabalance Response of extrabalance with 95% error bands, Finland, 1976Q1-1998Q4 reer extrabalance y- y* extrabalance extrabalance extrabalance Response of extrabalance with 95% error bands, Finland, 1999Q1-2013Q reer extrabalance 0 y- y* extrabalance 0 extrabalance extrabalance

22 Figure 3: Impulse response functions for Greece reer tradebalance Response of tradebalance with 95% error bands, Greece, 1981Q1-2000Q y- y* tradebalance Response of tradebalance with 95% error bands, Greece, 2001Q1-2013Q3 reer tradebalance y- y* tradebalance Response of intrabalance with 95% error bands, Greece, 1981Q1-2000Q4 reer intrabalance y- y* intrabalance intrabalance intrabalance Response of intrabalance with 95% error bands, Greece, 2001Q1-2013Q3 reer intrabalance y- y* intrabalance intrabalance intrabalance Response of extrabalance with 95% error bands, Greece, 1981Q1-2000Q4 reer extrabalance 0.04 y- y* extrabalance 0.04 extrabalance extrabalance Response of extrabalance with 95% error bands, Greece, 2001Q1-2013Q3 reer extrabalance y- y* extrabalance extrabalance extrabalance 21

23 Table A1. Testing for unit roots (ADF-tests) FI FR DE GR IT NL PT ES AT IE tradebalance I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(0)* I(1) I(0)** I(0)** I(1) I(1) I(1) I(1) y I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(0)* I(1) I(1) I(1) I(1) I(1) I(0)** I(0)** y world I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) reer world I(1) I(1) I(0)** I(0)*** I(0)** I(0)** I(1) I(1) I(0)* I(1) I(0)** I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) tradebalance I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)** I(1) y- y world I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)** I(0)** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(1) I(1) reer world I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** intrabalance I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(0)** I(0)* I(1) I(1) I(1) I(0)*** I(1) I(0)** y emu12 I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) reer emu12 I(1) I(1) I(1) I(0)*** I(0)* I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) intrabalance I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)** y- y emu12 I(0)*** I(0)** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(I) I(I) reer emu12 I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** extrabalance I(1) I(1) I(1) I(0)** I(1) I(1) I(0)* I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) y RoW I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) reer RoW I(1) I(1) I(0)** I(0)** I(0)** I(1) I(1) I(1) I(0)** I(1) I(0)** I(1) I(1) I(1) I(1) I(1) I(0)** I(0)* I(0)* I(0)* extrabalance I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)** y- y RoW I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)** I(0)** I(0)*** I(0)*** I(0)** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(1) I(1) reer RoW I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** I(0)*** Notes: On the left-hand side ADF-test with a constant and on the right-hand side ADF-test with a constant and a trend. All time series contain a unit root and all differences are stationary. If this does not hold, the font is red. denotes year-to-year changes. *, **, *** denote statistical significance at 10%, 5% and 1% levels. Abbreviations: See details from Table A2. 22

24 Table A2. Data description Series Description Frequency Source tradebalance tradebalance is year-to-year change of trade balance (goods net exports). We observe goods net exports (in US dollars) for a reporting-country, trade balance, from DOTS, when World is chosen as the partner-country. Same as below Same as below intrabalance intrabalance is year-to-year change of intra balance, which was calculated as follows: Firstly, we calculated trade balance (in US dollars) excluding services against the euro area (EMU-12 countries) from bilateral balances. Secondly, we took quarterly data on GDPs (in national currencies) and converted these to GDPs (in US dollars) using quarterly averages of daily data on exchange rates between national currencies and US dollar. Thirdly, we calculated trade balances as ratios to GDPs. Trade balance in US dollars: quarterly data 1960:Q1-2013:Q3; GDPs in national currency: quarterly data 1960:Q1-2013:Q3; Exchange rates: daily data except for Denmark, Sweden and UK quarterly data 1960:Q1-2013:Q3 Trade balance in US dollars: Direction of Trade Statistics (IMF); GDPs in national currency: National sources via Datastream, Eurostat (for Greece), International Financial Statistics (IMF) (for Netherlands, Portugal, Spain, Denmark and Sweden); Exchange rates: WM/Reuters via Datastream, International Financial Statistics (IMF) (for Denmark, Sweden and UK) extrabalance extrabalance is year-to-year change of extra balance, which was calculated as follows: Firstly, we calculated trade balance excluding services against the rest of the world by substracting the intra balance (in US dollars) from good net exports (in US dollars); that is extra balance equals trade balance minus intra balance. The second and third step as above. Same as above Same as above reer world reer EMU-12 reer world is year-to-year change in CPI-based real effective exchange rate with trade-weights (41 partner countries). reer EMU-12 is year-to-year change in CPI-based real effective exchange rate in which set of partner countries consist of EMU-12 countries with trade-weights. Quarterly averages of monthly data 1970:Q1-2013:Q3 Quarterly averages of monthly data 1970:Q1-2013:Q3 Bruegel Zsolt Darvas (Bruegel) reer RoW reer RoW is year-to-year change in CPI-based real effective Quarterly averages of exchange rate in which set of partner countries consist of 29 monthly data 1970:Q1- non-emu countries with trade-weights. 2013:Q3 Zsolt Darvas (Bruegel) y Domestic GDP per capita year-to-year changes Quarterly data on levels 1960:Q1-2013:Q3 y- y world y- y EMU-12 Difference between domestic GDP per capita year-to-year Domestic: same as above. change and GDP per capita year-to-year change in the world World economy: quarterly economy. data on growth rates 1969:Q1-2013:Q1 Difference between domestic GDP per capita year-to-year change and GDP per capita year-to-year change in the euro area (EMU-12 countries). The latter was calculated as follows: The pre-1996:q1 period: GDP-weighted average of EMU-12 GDP per capita year-to-year changes (different serie for every EMU countries as the reporting country is always excluded); Since then: Eurozone-12 GDP per capita year-toyear changes (same serie for all countries). Domestic: same as above. EMU-12: quarterly data on levels 1960:Q1-2013:Q3 International Financial Statistics (IMF), National sources via Datastream (for Austria, Finland and Ireland), Eurostat (pre-2011:q2 period for Greece) World economy: International Financial Statistics (IMF) The pre-1996:q1 period: International Financial Statistics (IMF), National sources via Datastream (for Austria, Finland and Ireland), Eurostat (pre-2011:q2 period for Greece); Since then: Eurostat Eurozone- 12 y- y RoW Difference between domestic GDP per capita year-to-year change Same as above and GDP per capita year-to-year change in the rest of the the world. The latter was calculated as follows: Let α be the share of the EMU- 12 countries of the world economy, y world the growth rate of the world economy, and y emu12 the growth rate of the EMU-12 countries. Then the following equality holds: y row =1/(1- α) y world +α/(α-1) y EMU-12. We applied this formula and used a constant weight, α, which is the sum of Austria, Finland, France, Germany, Greece, Italy, Netherlands, Portugal and Spain GDPs in 1995 divided by the World GDP in Data on GDP levels in 1995: World Development Indicators (World Bank). Other series same as above. 23