Putting Currency Misalignment into Gravity: The Currency Union Effect Reconsidered∗ Jan Hogrefe†

Benjamin Jung‡

Wilhelm Kohler§

ZEW Mannheim University of T¨ ubingen

University of T¨ ubingen CESifo

University of T¨ ubingen CESifo and GEP

August 2013

Abstract We argue that existing estimates of the trade effects of the euro are potentially biased because they ignore the macroeconomics of a common currency, which suggests potential misalignment from lack of nominal exchange rate adjustment in the face of asymmetric shocks. We develop an “augmented” gravity model of bilateral trade which allows us i) to provide consistentestimates of the real trade cost effect of the euro, irrespective of whether trade flows have been affected by euro-induced misalignment, and ii) to test for the presence of such misalignments and to consistently estimate their quantitative importance. Our empirical results confirm the negligibly low estimates for the trade cost effect of the euro obtained by recent literature, while at the same time finding strong support for misalignment-induced trade effects. These effects are vastly different across euro-zone member countries. We translate our econometric estimates into a picture of country-specific trade effects caused by misalignment over the period from 1999 to 2007.

JEL Codes: F12, F13, F15 Keywords: Euro, gravity model, real exchange rate, currency misalignment, unit labor costs ∗

We would like to thank workshop and conference participants at Hohenheim University, G¨ottingen University, ZEW Mannheim, the SMYE 2010, the Verein f¨ ur Socialpolitik 2010 annual conference, the FIW Vienna research conference 2010, the WHU CEUS 2011 workshop, and the Econometric Society European Meeting 2012 for valuable comments. Special thanks go to Erdal Yalcin and Ralf Fendel for detailed comments on earlier versions of this paper. The views expressed in this paper are solely those of the authors. All possible errors are, of course, ours as well. Earlier drafts of this paper circulated as ZEW Discussion Paper 10-023 and University of T¨ ubingen Working Papers in Economics and Finance No. 32. † Centre for European Economic Research (ZEW), L7 1, D-68161 Mannheim, Germany. Email: [email protected] ‡ University of T¨ ubingen, Mohlstrasse 36, D-72074 T¨ ubingen, Germany. Email: [email protected] §

Corresponding author: University of T¨ ubingen, Mohlstrasse 36, D-72074 T¨ ubingen, Germany. Email: [email protected]. Phone: +49/(0)7071/2976013. Fax: +49/(0)7071/295223

1

Introduction

Forming or joining a monetary union involves a trade-off between the benefit of more transparent and less costly transactions for intra-union trade and the potential cost from a loss of monetary policy autonomy. For the euro-zone, established in 1999, the consensus estimates for additional trade caused by the euro on the grounds of lower trade costs vary between 0 and 5 percent (see Baldwin, 2010; Silva and Tenreyro, 2010). These estimates squarely focus on the microeconomics of lower trade cost. In this paper, we argue that the macroeconomics of currency unions implies trade effects that potentially invalidate the estimates that have so far been obtained for the microeconomic channel. We propose a gravity-based empirical framework which allows us to disentangle and consistently estimate both types of trade effects from a common currency. Figure 1 plots nominal relative unit-labor-costs for OECD country pairs in 2002 and 2007, unadjusted for movements in nominal exchange rates and distinguishing between euro-zone and non-euro-zone country pairs. The data are normalized such that all relative unit-labor costs for 1999 are equal to 1. Moreover, we avoid redundancy be dropping mere inverses, arranging cost ratios in such a way that all values for 2002 are larger than 1.1 Any value deviating from 1 can be considered as a bilateral cost discrepancy, implicitly taking the values for 1999, the year when the euro-zone was established, as a benchmark. Two observations stick out. First, lots of bilateral cost divergencies had accumulated within the euro-zone already by 2002 (values above 1), and in the majority of cases they have further increased in magnitude through time (values above the 45o -line). Secondly, cost discrepancies are no less prevalent among non-eurozone countries than for members of the euro-zone, although the increasing time trend seems somewhat less pronounced.2 Let us define the bilateral real exchange rate as the relative unit-labor-cost adjusted by the bilateral nominal exchange rate. For non-euro-zone countries suitable adjustments in nominal exchange rates might have kept the real exchange rates at their 1999 values, but we cannot tell from the figure to what extent such adjustments did in fact occur. For euro-zone member countries, however, such an adjustment was clearly impossible since they use a common currency; any cost discrepancy thus indicates a change in the real exchange rate. But this need not constitute a misalignment. For instance, if for a certain country pair the underlying shock was a change in relative demand, then the observed divergence in unit-labor-cost may simply reflect a change in equilibrium relative prices. But if the divergence reflects a cross-country difference 1

Note that the full sample would include the ratios of German to Italian as well as the inverse ratio of Italian to German unit-labor cost. 2

The data are from the OECD “Unit Labour Costs - Annual Indicators”. Bilateral relative unit-laborcosts are the ratio between the individual countries unit-labor-costs which are, in turn, defined as total labor-cost over real output. Unit-labor-costs are for the entire economy to adequately reflect linkages between tradable and non-tradable sectors.

1

Figure 1: Bilateral unit-labor-cost divergence in OECD countries in wage setting or nominal wage rigidities, then a nominal exchange rate adjustment may be required to restore the equilibrium real exchange rate. If such an adjustment is ruled out because of the euro, we speak of a euro-induced real exchange rate misalignment, which is likely to affect bilateral trade. This is what we mean by macroeconomic trade implications of a currency union. In this paper, we develop an “augmented” gravity model of bilateral trade which allows us to do two things: First, to provide consistent estimates of the real trade cost effect of the euro, irrespective of whether trade flows have been affected by euro-induced misalignment. And secondly, to test for the presence of such misalignments and to consistently estimate their quantitative importance. We thus let trade data tell us about the extent to which the bilateral unit-labor-cost divergencies depicted in figure 1 have indeed constituted real exchange rate misalignments. And we let them tell us whether the euro was playing a causal role for such misalignments, if any. Our augmented gravity model allows for bilateral nominal cost divergencies to affect bilateral trade through limited exchange rate flexibility. Nominal exchange rates are modeled as reacting to such divergencies according to a reduced form equation which includes importer and exporter country fixed effects that capture these countries’ macroeconomic policies, as well as an indicator variable for common currency country pairs. The standard gravity model then emerges as a special case where nominal exchange rates consistently absorb all bilateral cost divergencies. We call the trade levels determined in this hypothetical case the “gravity norm”, and we define the trade cost effect of a common currency as shifting these gravity norm levels of trade. Actual trade levels emerge as misalignment-induced deviations from this norm. Importantly, our model

2

allows for such deviations to occur also for countries that do not belong to a currency union, acknowledging that real exchange rate misalignments may arise for reasons entirely unrelated to currency unions. Crucially, however, it allows us to test whether such misalignments are more likely, other things equal, for members of a currency union than for other countries. We estimate this model based on OECD data for the years 1995 to 2007. Our empirical results reaffirm the current consensus view of a negligibly small trade effect of the euro through the microeconomic (real trade cost) channel (see Baldwin et al., 2008; Baldwin, 2010; Silva and Tenreyro, 2010). More importantly, however, they support the hypothesis of euro-related currency misalignment. In other words, the macroeconomic channel for trade effects of the euro is empirically important. This result is in line with Berger and Nitsch (2013) who report longrun evidence of larger and more persistent trade imbalances in times of stricter exchange rate coordination.

2

An augmented gravity model

Suppose we have the usual Dixit-Stiglitz model of monopolistic competition serving as a theoretical underpinning of the gravity equation for bilateral trade volumes. The standard equilibrium conditions behind this equation determine all factor prices in all countries as functions of countries’ factor endowments, their technologies, and bilateral real trade costs (including the effects of currency unions according to the transaction cost channel). Goods prices in each country then follow from markups as well as real trade costs. Since all equilibrium conditions are homogeneous of degree zero in prices, factor prices are determined only up to a common num´eraire. If we assume, for simplicity, that labor is the only primary input, then the model endogenously determines the relative wage rates between all countries considered. Obviously, with N countries there are N − 1 independent relative wage rates to determine. The real wage within each country then follows from goods prices feeding into the exact price index of each country. Given a uniform technology across all countries, and given a certain pattern of bilateral real trade costs, the home market effect implies that country size confers an advantage in terms of a higher wage rate; see Krugman (1980). Hence, differential growth in the labor force entails changes in relative wage rates. Note that the model leaves nominal wages, expressed in national currencies, as well as nominal exchange rates undetermined. Thus, the formation of a currency union by a subset of countries does not appear to make any difference, beyond reducing trade costs. However, this changes if mechanisms of wage setting impose binding restrictions on nominal wages. For instance, nominal wages might be set before labor productivity is revealed, and fixed with downward rigidity for some time thereafter, so that equilibrium relative wages imply certain values of bilateral nominal exchange rates. Indeed, if labor market institutions impose N binding

3

restrictions on nominal wages, then the model determines N − 1 independent equilibrium values of nominal exchange rates. Whenever a currency union among certain countries makes such an equilibrium adjustment of all nominal exchange rates impossible, relative wage rates will deviate from their equilibrium values. We call this a misalignment due to the currency union. It is obvious that misalignment of wage rates and currencies are two sides of the same coin. The outcome is a multiple deviation from “gravity norm” levels of bilateral trade, potentially coupled with unemployment in countries with an overvalued currency. Denoting the c.i.f.-price in country j for a variety arriving from country i by pij , the quantity of demand Dij for this variety in country j is Dij = Aj (Pj )σ−1 (pij )−σ ,

(1)

where σ > 1 denotes a uniform elasticity of substitution between different varieties of goods. In this expression, Aj is equal to country j’s expenditure, and Pj is the exact price index (unit-expenditure function), depending on prices of all varieties shipped to market j: Pj =

hX i

Ni (pij )1−σ

i1/(1−σ)

,

(2)

where Ni is the number of varieties produced by country i. Importantly, all variables on the right-hand side of (1) are in country j’s currency. Following standard literature we use Tij to denote “iceberg-type” real trade cost. Using ai to denote country i’s variable labor input requirement per unit of output, wi to denote country i’s nominal wage rate and Eij to denote the price of its currency in units of currency j, the c.i.f.-price of a typical variety exported from i to j, measured in country j 0 s currency, is equal to pij = Eij

Tij wi ai , ρ

with ρ := (σ − 1)/ σ < 1.

(3)

Writing Xij = Ni Dij pij for the overall value of bilateral exports and using equations (1) and (3), and expressing trade flows in units of the num´eraire currency 1, we have σ−1

Xij Ej1 = ρ

σ−1

Ni Aj Ej1 (Pj /wj )

1−σ

(Tij ai )

  wi 1−σ Eij . wj

(4)

Note that this expression still has endogenous wage rates on the right-hand side. However, the goods and labor market equilibrium conditions may be used to get rid of these wage rates, so that Xij , relative to the product of the two countries’s GDPs, depends on exogenous bilateral trade costs and on the two countries’ multilateral resistance (Anderson and Van Wincoop, 2003). The challenge now is to disentangle the above mentioned two channels. Through the microeconomic channel, a currency union between countries i and j affects their bilateral trade trade via Tij .

4

Moreover, Xij is also affected by currency unions between countries other than i and j through the aforementioned multilateral resistance terms. Through the macroeconomic channel, trade between i and j is affected by deviations of Ej1 as well as Pj /wj and Eij wi /wj from their “gravity norm” values, i.e., by misalignment. Moreover, by complete analogy to multilateral resistance, trade between i and j is also affected by misalignment of relative wage rates between countries other than i and j through the price index Pj . Proposition 1 (microeconomic channel) Irrespective of whether or not there is misalignment, all parameters that drive the microeconomic channel of real trade costs Tij , including the transaction cost effect of a common currency, may be consistently estimated by regressing bilateral exports on the relevant trade barrier controls behind Tij , provided that the estimation includes time-varying country fixed effects that distinguish between a country’s role as an exporter and importer, respectively. Proof. Currency arbitrage implies Eij = Ei1 /Ej1 . Inserting into equation (4) we realize that potential misalignment is fully controlled for by fixed effects as stated in the proposition. Note that the presence of misalignment requires role-specific and time-varying fixed effects even if trade costs Tij are symmetric, as often assumed in the literature. In essence, the macroeconomic channel depends on whether restrictions on nominal wages deriving from countries’ labor market institutions become binding at the ongoing nominal exchange rate. If they don’t, then this channel is dead. If they do in countries with separate currencies and flexible exchange rates, then nominal exchange rate adjustments may restore equilibrium relative wage rates. But for countries that are bound together by a common currency such an adjustment is ruled out, which gives rise to a misalignment-induced deviation of trade flows from their “gravity norm”. According to this view, nominal exchange rates Eij and nominal factor prices wi and wj are determined jointly though an interaction between the monetary policy regimes in countries i and j, given these countries’ labor market institutions governing nominal wage setting. Our maintained hypothesis about the macroeconomic channel now holds that, controlling for other country-specific effects, this interaction is different for countries that belong to a currency union from those that don’t. We abstain from developing a structural model of exchange rate determination, but simply represent this interaction through the following reduced form equation: Eij = (mij )αij η i η j .

(5)

In this equation mij measures relative unit labor cost, wi /wj , relative to some benchmark value, say at the beginning of the currency union in question. In our case this is 1999, the year when the EMU was formed. The two terms η i and η j are exporter- and importer-country-specific

5

effects capturing, among other things, the aforementioned institutions responsible for nominal wage restrictions. With C currencies there can be no more than C − 1 independent nominal exchange rates Eij . Hence, the parameter αij cannot vary freely across all pairs. To proceed towards an estimable framework, we formulate the following hypothesis: αij =

α0 α0 + α1

for non-euro-zone country pairs for euro-zone country pairs.

(6)

The parameter α0 captures the relationship between bilateral cost-divergence and the bilateral nominal exchange rate for countries that have their own currencies. The parameter α1 takes care of the fact that Eij is restricted to a value of 1, if countries i and j share the same currency.3 Since our prime interest lies in trade effects of the EMU, we do not estimate (5), but instead insert it into (4), the augmented gravity equation for trade flows. More specifically, we replace (Eij mij )1−σ = (mij )β 0 +β 1 eij η i η j , were eij = 1 indicates member countries of the currency union and eij = 0 indicates otherwise. The estimated coefficient β 0 = (1 − σ)(1 + α0 ) measures the extent to which bilateral cost divergencies affect bilateral trade for countries with different currencies, while the coefficient β 1 = (1−σ)α1 measures the differential effect caused by such cost divergencies, if the countries in question are members of a currency union. The interpretation of α0 and α1 is as follows. If α0 = −1, then for non-euro country pairs with separate currencies any cost divergence is completely offset by a nominal exchange rate adjustment, leaving bilateral trade unaffected. If at the same time α1 = 0, then cost divergence for euro country pairs similarly leaves trade unaffected, meaning that movements of nominal wage rates as required according to the “gravity norm” have not met any binding constraint imposed by labor market institutions. If α0 > −1, then changes in relative nominal wage rates are less than fully compensated for by nominal exchange rate changes even for countries with separate currencies, so that trade is affected by relative wage rate misalignment. And if α1 > 0, then the resulting misalignment is aggravated for country pairs that belong to the euro-zone, since this rules out any compensating nominal exchange rate adjustment. Proposition 2 (macroeconomic channel) If the estimated coefficients of the augmented gravity model satisfy the condition β 0 + β 1 < β 0 < 0, then bilateral cost divergencies cause misalignment of relative wage rates also for countries with separate currencies, but the resulting effect on bilateral trade is stronger for countries who share the same currency. Proof. Since σ > 1 by assumption, β 0 < 0 implies α0 > −1 and β 1 < 0 implies α1 > 0. Moreover, β 0 + β 1 < 0 implies α0 + a1 > −1. Note that our model allows for misalignments also 3

Equation (5) is not intended as a causality statement. Causality can go both ways. For instance, with respect to the EMU, it is probably fair to say that in pre-euro-times mij has driven Eij , and there was a hope that in euro-times we would have Eij = 1 driving mij .

6

among countries that do not belong to a currency union.

3

Empirical estimation

The baseline specification of our estimation equation is ln

Xijt Ej1t = γeijt + β 0 mijt + β 1 (eijt × mijt ) + %RTAijt + ν ij + ν it + ν jt + uijt , Yit Ei1t Ajt Ej1t

(7)

where RTAijt measures common membership in a regional trade agreement, and ν ij , ν it and ν jt denote pair-specific fixed effects, exporter×time effects and importer×time effects, respectively, while uijt denotes an error term which is assumed to be iid. This equation follows directly from (4), after substitution of (5) and (6) and realizing that all time-invariant dyad-specific gravity controls (distance etc.) are captured by ν ij , while all time-varying importer- and exporterspecific controls are nested in ν it and ν jt . Note that the trade ratio on the left normalizes by the product of exporter GDP and importer absorption, Aj (instead of GDP Yj ), thus taking into account current account imbalances as implied by misalignments. Exporter GDP satisfies Yi = Ni pi yi , where under standard assumptions output per firm yi is given parametrically by technology. Division by Yi in equation (7) thus implies that the term pi yi gets picked up by the exporter×time effect ν it . Equation (7) is estimated on first differences of the data, whereby the interaction term is specified as β 1 (eijt × ∆mijt ). First-differencing the data controls for all time-invariant pairspecific influences on bilateral trade. As such, it also helps in controlling for potential selection of countries into euro-zone membership based on past trade levels (Baier and Bergstrand, 2007).4 Our estimation relies on a panel of 20 OECD countries for the years from 1995 through 2007.5 Merchandize trade data are taken from the IMF Direction of Trade Statistics (DoTS), gravity controls including GDPs are from CEPII, and current account data (used to compute Aj ) are from the OECD. Information on RTAijt is from the WTO. Relative nominal cost conditions, wi /wj , are measured using economy-wide unit-labor-cost from the OECD, as already described in section 1; for a robustness check we also use manufacturing labor-cost. 4

A further potential reason for endogenous selection into euro-zone membership could be due to optimum currency area considerations, so that countries joining the union are structurally more similar than others. This possibility is discussed and forcefully rejected in Wyplosz (2006). 5

We aim for a sample which is similar to those of earlier studies (Micco et al., 2003; Silva and Tenreyro, 2010). The countries included are: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, Korea, Netherlands, Norway, Portugal, Spain, Sweden, United Kingdom, and United States. The time span was chosen with data quality considerations in mind. In particular, Baldwin (2006) discusses statistical weaknesses of pre-1993 trade data for European countries. We drop the years 1993 and 1994 to avoid having to control for the effect of EU accession of some countries - a choice that does not affect our results except for adding clarity in the exposition.

7

Our primary interest lies in the first three terms on the right hand side of (7). The estimated coefficient γˆ measures the trade effect of the euro through the trade cost channel.6 From proposition 1, we know that this can be estimated consistently without bothering about euro-induced misalignment, i.e., setting β 0 = β 1 = 0, provided we allow for role-specific country×time effects ν it and ν jt . This is what we do in column 1 of panel (c) in table 1, which clearly is the preferred specification. To allow for comparison with earlier studies, we include panels (a) and (b), although these must be regarded as misspecifications since they do not include role-specific fixed effects. The results clearly corroborate the literature consensus of a negligibly small trade cost effect of the euro; none of the estimated coefficients is economically large and/or statistically significant. Somewhat surprisingly, the RTA-effect, if significant, is negative, which is at odds with other estimates in the literature (Baier and Bergstrand, 2007). Our explanation for this is that first differencing the data leaves but a single RTA with time-varying membership between 1993 and 2007, which is the Australia-US free trade agreement. Columns (2), (3) and (4) of table 1 estimate the full model, additionally allowing for trade effects from euro-related misalignment by freeing both β 0 and β 1 from zero. Unfortunately, proposition 1 now turns against us, since it implies that the misalignment term mijt is composed of an exporter×time and an importer×time effect. Hence, given the presence of ν it and ν jt in (7), the parameter β 0 cannot be identified in this specification. Identification of β 0 is possible for specifications (a) and (b), but we repeat that (c) clearly is the preferred specification. The estimates in column (2) support our hypothesis that for trade between member countries a loss in bilateral cost competitiveness feeds into a stronger negative trade effect than for nonmembers. The effect is largest in our preferred specification. To test for a selection bias, we ask whether the euro area members already had a different responsiveness to cost divergence in the years leading up to the currency union. We do so by setting eijt = 1 for eventual euro-zone ˆ for this country pairs already for t < 1999 (t < 2001 for Greece). The coefficient estimate β 1

coding is labeled “pre-euro-induced misalignment” in tables 1 and 2. Looking at columns (3) and (4) of table 1, statistically significant estimates appear only for specification (b), which is a misspecification. In our preferred specification (c), this effect is clearly absent. Although panels (a) and (b) seem to suggest that for trade with or between non-member countries cost divergence is fully absorbed by nominal exchange rates, we are hesitant to push this interpretation. In all panel (a) and (b) regressions, unobserved country heterogeneity remains uncontrolled for due to the less comprehensive structure of the fixed effects. Table 2 reveals that all major results are robust to a somewhat more narrow definition of unit-labor-cost covering only the manufacturing sector. 6

The estimated euro-induced percentage increase in the trade ration is given by [exp(ˆ γ ) − 1] × 100 percent.

8

4

A picture of country heterogeneity

While the trade cost reductions made possible by the euro very likely are of the same magnitude for all euro-zone member countries, the euro-induced misalignment is vastly different across ˆ allows us to paint a picture of country-specific trade countries. The estimated coefficient β 1 ˆ = effects that derive from such misalignment. We do so by using the coefficient value β 1

−0.527 from column 2 of panel (c) in table 1, in order to calculate the following country-specific cumulative trade effect: X ∆xij,07−99 = xij,99

j∈euro

  x ˆ (mij,07 − mij,99 ) . P ij,07 ×β 1 xij,07 mij,99

(8)

j∈euro

On the right-hand side of this equation, mij,99 and mij,07 are sample values of our cost-discrepancy measure at the beginning of the euro-zone in 1999 and the end-year of our sample which is 2007. xij,07 is the value of bilateral trade normalized by exporter GDP and importer absorption. Bilateral cost discrepancies are thus weighted by partner country trade shares. The results are found in table 3. Given the large differences in cost developments across member states, we naturally see pronounced variation in the country specific trade effects. In particular, Germany and, to some degree, Austria and Finland have seen positive trade effects from the euro via the misalignment channel. Exports from Belgium and France are only marginally affected, since their relative unit labor costs have remained fairly balanced towards the rest of the euro area. However, the southern member states as well as Ireland have seen their unfavorable unit labor cost developments translated into negative trade effects.

5

Conclusion

We conclude with four remarks. First, countries considering to join the euro, or any other currency union, should not expect a sizable and balanced increase in their exports and imports to and from other union members. The conventional trade cost channel does not appear to deliver a significant trade effect from using a common currency. Moreover, there is a second channel that has so far been neglected in the literature. If countries feature different wage setting institutions and/or different degrees of nominal wage rigidities, then tying them together by the nominal anchor of a common currency is likely to cause real exchange rate misalignment, which in turn affects their bilateral trade. For optimum currency areas, this effect should be zero. Estimation of a suitably specified gravity model on the basis of an OECD data set reveals that the euro-zone is no such optimum currency area, since we do observe significant trade effects through this misalignment channel. However, this channel almost by definition delivers vastly different trade effects across dif-

9

ferent countries. Even if a certain country expects to remain in line with the nominal anchor which is set by the union monetary policy, it will still be affected by other countries’ inability, or unwillingness, to do the same. A case in point is Germany where our gravity-based estimate implies a misalignment-induced increase in exports to euro-zone members, which totals to about 8.5 percent for the period from 1999 to 2007. This is mirrored by negative effects of around 4 percent for Spain and Greece, and as much as 12.5 percent for Ireland. It is hard to imagine that adopting a common currency should have failed altogether in lowering real trade costs. But according to our estimates, lower trade costs have failed to spur much additional trade between euro-zone countries. If this interpretation is correct, then we must not conclude from a zero trade effect that the euro has failed to deliver gains from less costly trade. Even if positive, the trade effects would be the wrong place to look at when measuring such gains, since they focus only on the “triangular” effect. The bulk of welfare gains arise, not in the form of “triangular” gains, but in the form of first order “rectangular” gains from less costly transactions in existing trade volumes. And finally, when interpreting misalignment-induced trade effects of the euro, countries should avoid falling victim to mercantilist thought suggesting that more (less) exports is a good (bad) thing per se. A proper welfare interpretation of such trade effects requires an intertemporal perspective. If we nonetheless venture a welfare conclusion from a static perspective, then if anything the opposite is true: Any misalignment effect that boosts exports is akin to a deterioration of the terms of trade.

10

References Anderson, J. and Van Wincoop, E. (2003). Gravity with gravitas: a solution to the border puzzle. American Economic Review, 93(1):170–192. Baier, S. and Bergstrand, J. (2007). Do free trade agreements actually increase members’ international trade? Journal of International Economics, 71(1):72–95. Baldwin, R. (2006). In or out: does it matter?: an evidence-based analysis of the Euro’s trade effects. Centre for Economic Policy Research. Baldwin, R. (2010). The euro’s impact on trade and foreign direct investment. In Buti, M., D. S. G. V. and Martins, J., editors, The Euro - The First Decade, pages 678–709. Cambridge University Press. Baldwin, R., DiNino, V., Fontagn´e, L., De Santis, R. A., and Taglioni, D. (2008). Study on the impact of the euro on trade and foreign direct investment. European Commission Economic Papers, 321. Berger, H. and Nitsch, V. (2013). Wearing corset, losing shape: The euro’s effect on trade imbalances. Journal of Policy Modeling. forthcoming. Krugman, P. (1980). Scale economies, product differentiation, and the pattern of trade. American Economic Review, 70(5):950–959. Micco, A., Stein, E., and Ordo˜ nez, G. (2003). The currency union effect on trade: early evidence from EMU. Economic Policy, 18(37):315–356. Silva, J. and Tenreyro, S. (2010). Currency unions in prospect and retrospect. Annual Review of Economics, 2:51–74. Wyplosz, C. (2006). European Monetary Union: the dark sides of a major success. Economic Policy, 21(46):207–261.

11

Table 1: Trade cost and misalignment channel (economy-wide unit-labor-cost) (a): Time effects

(1)

Bilateral cost divergence (β 0 ) Euro-induced misalignment (β 1 )

(2)

(3)

(4)

-0.0150 (-0.12) -0.411** (-2.00)

-0.0254 (-0.22)

0.0500 (0.38) -0.466** (-2.20) -0.517 (-1.63) 0.0180 (0.75) -0.0801** (-2.17)

Pre-euro-induced misalignment Both in euro zone (γ) Both in RTA (%) N R2 (b): Country×time effects

0.0180 (0.75) -0.0801** (-2.19)

0.0180 (0.75) -0.0801** (-2.19)

-0.452 (-1.46) 0.0180 (0.75) -0.0801** (-2.20)

4560 0.167

4560 0.168

4560 0.168

4560 0.169

(1)

(2)

(3)

(4)

-0.0150 (-0.14) -0.411** (-2.15)

-0.0254 (-0.25)

0.0500 (0.43) -0.466** (-2.37) -0.517** (-1.97) -0.0181 (-0.47) -0.0664 (-1.23)

Bilateral cost divergence (β 0 ) Euro-induced misalignment (β 1 ) Pre-euro-induced misalignment Both in euro zone (γ) Both in RTA (%) N R2 (c): ν it and ν jt

-0.0181 (-0.47) -0.0664 (-1.23)

-0.0181 (-0.47) -0.0664 (-1.23)

-0.452* (-1.77) -0.0181 (-0.47) -0.0664 (-1.23)

4560 0.301

4560 0.302

4560 0.302

4560 0.303

(1)

(2)

(3)

(4) -0.526* (-1.78) -0.110 (-0.33) -0.0181 (-0.48) -0.0664 (-0.67) 4560 0.393

Euro-induced misalignment (β 1 )

-0.527* (-1.78)

Pre-euro-induced misalignment Both in euro zone (γ) Both in RTA (%) N R2

-0.0181 (-0.48) -0.0664 (-0.67)

-0.0181 (-0.48) -0.0664 (-0.67)

-0.113 (-0.34) -0.0181 (-0.48) -0.0664 (-0.67)

4560 0.392

4560 0.393

4560 0.392

Estimation method: First differences. Dependent variable: Bilateral exports scaled by exporter’s GDP and importer’s expenditure. Cluster-robust t-statistics are in parentheses. ∗∗ and ∗ indicate significance at 5% and 10%, respectively. Fixed effects are not shown.

12

Table 2: Trade cost and misalignment channel (manufacturing unit-labor-cost) Specification: ν it and ν jt

(1)

Euro-induced misalignment (β 1 )

(2)

Both in RTA (%) N R2

(4) -0.364* (-1.65) -0.184 (-0.62) 0.0127 (0.36) -0.0583 (-0.76) 4256 0.408

-0.376* (-1.71)

Pre-euro-induced misalignment Both in euro zone (γ)

(3)

0.0127 (0.36) -0.0583 (-0.76)

0.0127 (0.36) -0.0583 (-0.76)

-0.228 (-0.77) 0.0127 (0.36) -0.0583 (-0.76)

4256 0.408

4256 0.408

4256 0.408

Estimation method: First Differences. The dependent variable is bilateral exports scaled by exporter’s GDP and importer’s expenditure. Cluster-robust t-statistics are in parentheses. ∗ indicates significance at 10%. All regressions include role-specific exporter-and-time effects (not shown).

Table 3: Country specific effects of misalignment Country Germany Austria Finland Belgium France Portugal Italy Netherlands Spain Greece Ireland

Change in exports in % using βˆ 1 = −0.527

change in mij

8.46 3.20 2.32 0.30 -0.24 -1.74 -2.34 -2.73 -4.15 -4.19 -12.53

-16.05 -6.08 -4.39 -0.58 0.45 3.30 4.43 5.18 7.87 7.95 23.77

The values show changes in exports due to the misalignment channel. They are based on a multiplication of the estimated coefficient from table 1, column 2 in panel (c) with the observed change in weighted misalignment, measured through unit-laborcosts, between 1999 and 2007.

13