The introduction of the Euro and trade imbalances in the Eurozone

(1)

University of Amsterdam, August 2015

The Introduction of the Euro and Trade Imbalances in the Eurozone

Master’s Thesis Economics

Track: International Economics and Globalization

Author:

Joep van der Bijl

Student number:

6114075

E-mail:

joepvdbijl@gmail.com

Supervisor:

Dr. John Lorié

Second Reader:

Dr. Dirk Veestraeten

(2)

1. Introduction

On January 1, 1999, eleven countries abandoned their own national currencies and adopted the Euro as their new currency. This group was later extended by an additional seven members. One of the main reasons for establishing the Economic and Monetary Union (EMU), next to the argument for political integration, was to foster further trade (CBS, 2011).

In the last decade, especially the starting years of the EMU, there has been a growing number of studies that empirically examine the impact of the introduction of the euro on international trade. Micco, Stein and Ordoñez (2003) among others find that for pairs of countries that adopted the euro, international trade has increased by about 5 - 10 percentage points and intra Eurozone trade by 9 - 20 percentage points. Bun and Klaassen (2006) later extended the standard model by including pair-specific time trends and this resulted in a considerably lower estimated euro impact of only 3 percentage points. Accordingly, most other studies broadly find a positive significant increase in trade as a result of the adoption of the euro, however, their findings only vary in the magnitude of this effect (Rose, 2000; Faruqee, 2004; Baldwin, 2006).

Figure 1: Evolution of trade imbalances of the EU-15 countries from 1980 till 2012.1

Source: Own calculations based on data from the IMF Direction of Trade Statistics (IMF DOTS, 2012)

This study, however, will not focus much on the impact of the euro on total trade, but rather its impact on the dispersion of trade. In fact, an important fear about the EMU was that

1_{Overall trade balances are given in terms of the trade balance per country with all other EU-15 countries as a}

percentage of total trade (with these countries). The full derivation is available in the data file (Excel).

-0,8 -0,6 -0,4 -0,2 0 0,2 0,4 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

Intra EU-15 trade balances (% of total trade)

AUS FIN FRA GER GRE IRE ITA NED POR SPA DEN SWE UK SWI NOR

(4)

- 3 -

specific shocks, in the absence of national currencies, result in larger and more persistent trade imbalances between member states (Feldstein, 1997; Decressin and Stavrev, 2009). Current experiences have shown that fears have become truth. While the external balance of the Eurozone as a whole has remained close to balance, several individual countries have experienced sizeable surpluses or deficits: Spain, Greece and Portugal ran large deficits by historical norms for industrial economies, while Germany and other smaller countries such as the Netherlands, Belgium, Austria and Finland ran large surpluses. (Schmitz and von Hagen, 2011; Berger and Nitsch, 2010). Figure 1 depicts the trend in intra euro-area trade balances and supports this finding by showing a small increase in the spread of individual countries’ trade balances since the start of the European and Economic Monetary Union (EMU) in 1999.

Figure 2: Trend in the absolute value of the average bilateral trade balances. 2

Source: Own calculations based on data from the IMF Direction of Trade Statistics (IMF DOTS, 2012)

This divergence in trade is even more obviously displayed by figure 2 that depicts the trend in the average magnitude of the bilateral trade balances among the same EU-15 members. Although the figure shows a boost around 1999, the increase in trade imbalances seems to be a trend that already started since the late 80’s. The decline in trade imbalances since the late 00’s represents external adjustments of some south-European deficit countries since the start of the crisis in 2008.3

However, most of these improvements in the trade balance seem to be the result of income effects; crisis drove down income, income drove down imports, while exports remained relatively more stable (Sinn and Valentinyi, 2013). A natural question then is whether these imbalances can be

2_{The average magnitude of the bilateral trade balance (among EU-15 members) is derived by taking the}

average of all 15 trade balances (as a percentage of total trade), used in figure 1, in absolute values. The full derivation of the figures is available in the data file (Excel).

3_{Giavazzi and Spaventa (2010) even attribute the crisis in the Eurozone to the accumulation of imbalances in}

deficit countries.Timing and scales of adjustments in these countries were heterogeneous. Spain and Ireland large improvements (6 % of GDP), Portugal and Greece somewhat smaller, and only in Italy the trade balance deteriorated (Sinn and Valentinyi, 2013)

0,06 0,08 0,1 0,12 0,14 0,16 0,18

(5)

- 4 -

explained by fundamental factors such as EMU-membership itself, which will also be central in this paper. In this regard, it is important to analyze whether and how much the euro contributes to the dispersion of trade, and if so, through which channels. Given the fact that there exist a huge literature studying the impact of the euro on total trade flows, it is rather surprising that such an important issue has not been carefully addressed in the empirical literature yet.

Economic theory suggests that the abolition of nominal exchange rate volatility – the case of a common currency – should increase bilateral trade imbalances. The absence of flexible exchange rates poses a particular problem within a monetary union that also operates a single goods and financial market, such as the euro area. In this case, prices need to change in order to restore the external account, but as prices are in fact highly inflexible this will cause more persistent trade imbalances (Friedman, 1953; Berger and Nitsch, 2010). The picture, however, remains complicated due to the historically high propensity of European countries to trade with each other; a trend already facilitated by, among others, the creation of the internal market within the EU.4_{In the same}

way, the European Commission (2009, 2010) observes that bilateral trade imbalances among euro area member countries have diverged and that this seems to have been on an upward trend already prior to the introduction of the euro, starting in the early 1990s and reaching his peak in 2008. This has led to the following research question:

“What is the effect of the introduction of the euro on trade imbalances within the Eurozone?”

Although there are a number of empirical studies (Jaumotte and Sodsriwiboon, 2010; Schmitz and von Hagen, 2011; Chen et al., 2013; Nieminen, 2014) that attempt to identify the main determinants of trade- and current account balances, there is – up to my knowledge – only one study by Berger and Nitsch (2010) that directly examines the impact of the euro on the regional European trade imbalances. By employing bilateral data on 18 EU member countries over the period 1948-2008 they find that, as a result of the introduction of the euro, trade imbalances among the euro area members widened by about 2-3 percentage points. However, as I will discuss later in this paper, I think their model is thin and incomplete. Their model only considers a euro-impact and does not comprehensively control for the possible prevalence of the before mentioned upward trends, that are actually caused by other factors. One of the implications of the omitted variable bias is then that

4_{The creation of the EMU proceeded in three stages. The first consisted of financial liberalization in order to}

complete the single market and broadly took place between 1990 and 1993 (European Union). The second stage was marked by economic policy convergence and reduced exchange rate volatility (through the Exchange Rate Mechanism) in preparation for the introduction of a common currency. The final stage was the introduction of the euro which took place in 1999 (Jaumotte and Sodsriwiboon, 2010).

(6)

- 5 -

the euro-effect could be biased upwards as it will capture general economic integration among the member states, and not merely the currency effect.

Hence, in a first exercise I reproduce the standard model of Berger and Nitsch (2010) – with only a euro dummy and conventional types of fixed effects – and comprehensively examine the direct euro-effect on trade imbalances from various angles (e.g. with and without linear and pair-specific trends, samples with and without the crisis years). Subsequently, I compare the results with those of similar regression specifications carried out by Berger and Nitsch (2010). In a second exercise, I examine the euro-impact in more detail by augmenting the derived ‘standard’ model with important determinants of trade imbalances as proposed by the theoretical literature. In fact, the theoretical literature explains the rising trade imbalances largely as results of other driving forces, such as disparities in relative income, competitiveness and the way governments conduct their macroeconomic policies, from which I take fiscal imbalances as the indicator. Moreover, as the next sections will describe, the impact of some of these driving forces may have been intensified by the introduction of the common currency and its associated institutional framework (Belke and Dreger, 2013; Eichengreen, 2010). Hence, by letting the additional explanatory variables interact with the euro-dummy in a final exercise, one can get a clearer understanding of the mechanisms through which the suggested trade imbalances under the euro area may have arisen.

In order to derive the euro-effect on trade imbalances effectively, absolute values of the bilateral trade balances are regressed using Ordinary Leased Squares (OLS) as the estimation strategy. The sample covers a time period from 1980 till 2011 and comprises a homogeneous set of 15 European countries, 10 of which have adopted the euro as their common currency.5_{In line with}

previous work of Berger and Nitsch (2010, 2008), the approach of picking these group of countries has the advantage of including countries which either share the European Union’s institutional framework or are closely associated with it.6

This paper will contribute to the existing literature in several ways. First, most research that has been done focuses on a broader definition of imbalances, like current account imbalances in general, and not particularly on trade imbalances of the developed European countries.7_{Second, in}

the debate of global imbalances the imbalances between the USA and some Asian countries, among others, are mostly examined whereas the Eurozone did not receive much attention until recently (Belke and Dreger, 2013). Finally, the study in general is important for policymakers as it affects the

5_{13 countries constituting the EU from the start (the EU-15 excluding Luxembourg and Belgium due to data}

availability) plus Norway and Switzerland. EU member countries (countries in black adopted the euro):

Germany, Greece, Spain, Italy, Netherlands, France, Ireland, Austria, Portugal, Finland, Denmark, Sweden and

Great Britain.

6_{Most of the countries are now exposed to large amounts of EU-related integration.}

7_{Specifically, data on bilateral trade, provided by the IMF DOTS, is on the value of merchandise exports and}

(7)

- 6 -

debate in non-euro countries on whether to join the euro or not. In the incumbent euro countries, on the other hand, it remains relevant for policymakers to keep track of the trade costs and benefits of adopting a common currency.

The remainder of the thesis is structured as follows. The next section provides a summary of the literature and consists of two parts; a theoretical and an empirical part. The first part describes the theoretical mechanisms through which a common currency could contribute to larger mutual trade imbalances and is meant to get a clearer understanding of the economic determinants of trade imbalances. The second part summarizes the most important empirical studies in this field of research thus far. On the basis of both parts I introduce and discuss my own model in section 3, where I give a detailed description of, among others, the included variables, the relevant fixed effects and trends, and the raw data. In section 4 the estimation results are presented and discussed, both for the basic model with only a euro-dummy and for the augmented models with additional explanatory variables and interaction terms. The final section concludes.

(8)

- 7 -

2. Trade imbalances in the euro area

This chapter is about the existing literature and consists of two parts. The first subsection will describe the theoretical genesis of trade imbalances when countries opt for a common currency. By doing this, one can get a better understanding of the mechanisms through which these imbalances are built up. The second subsection will revise the most relevant empirical findings, for this study, thus far. By means of a detailed notion of both the theoretical and empirical literature I will be able to derive an extensive model, which I will discuss and estimate in sections 3 and 4.

Although the trade balance will be the focus of this study, much of the current account literature will be taken into account since the theoretical and empirical literature on trade imbalances is thin. Besides, the trade account is the most important part of the current account, so that developments of the current account are largely explained by developments of the trade balance (Mankiw, 2010; Schmitz and von Hagen, 2011). 8

2.1 Emergence of trade imbalances: Catching up, competitiveness or other drivers?

According to the literature, current account and trade balance developments in the Eurozone might be traced back to both catching up and competitiveness factors. The catching-up component implies that external accounts might reflect a healthy convergence process between countries with different levels of income per capita. The second explanation consists of several factors determining competitiveness and is more straightforward – relatively more competitive countries should rather run bilateral trade surpluses than less competitive countries. Both mechanisms will be explained comprehensively in the next subsections. In addition, the potential impact of macroeconomic policy differences on trade imbalances in the euro area will be discussed. Particularly, the emergence of fiscal imbalances will get the attention. Besides the fact that government spending has a direct impact on the trade balance via the ‘twin-deficit hypothesis’, it has also become the major macroeconomic policy tool in situations like the Eurozone where monetary policy is centralized.

2.1.1 Catching up

The catching up component implies that external accounts, such as the trade balance, might reflect a healthy convergence process between counties with different levels of per capita income. The underlying theory is based on the assumption that if investors are able to easily invest anywhere in the world – acting rationally – they would invest in countries that offer the highest return per unit on investment. This would drive up the price of investment until returns across the countries are similar.

(9)

- 8 -

In line with the intertemporal approach of the current account, explained in more detail by Lane and Pels (2011), countries with lower levels of per capita income attract foreign capital as these countries have higher growth prospects where the rates of return on investment are higher. They will borrow money abroad to finance their lagged development by increasing domestic investments (Blanchard and Giavazzi, 2002; Lane and Pels, 2011). At the same time, the higher growth prospects of the poorer countries will cause agents to anticipate higher permanent income. Hence, as domestic agents, such as firms and households, want to smooth consumption over time, higher permanent income leads to higher consumption and less savings today. Investments are expected to exceed savings, inducing net capital inflows and external deficits in the catching up period. Symmetrically, richer countries should run larger external surpluses during the same period since they save more than they invest. According to the catching-theory, this process ends when investment rates in both countries are equal again (Blanchard and Giavazzi, 2002).

In this sense, the catching up period – where capital flows from relatively rich to relatively poorer countries – is, under appropriate macroeconomic conditions, part of a healthy convergence pattern between economies. The observed current account deficits of the southern Eurozone countries should in this view be regarded as signs of the proper functioning of the European Monetary Union rather than signs of inappropriate macro-economic policies (Schmitz and von Hagen, 2011).9_A

balanced trade balance is not optimal in the short run and policy measures directed to restore the imbalances might be even more harmful (Belke and Dreger, 2013).

Although empirical research established that the “catching-up theory” does not hold on a global level, Schmitz and von Hagen (2011) and Blanchard and Giavazzi (2002), among others, show that capital does flow from rich to poor within Europe. They provide evidence that the net capital flows follow differences in per-capita income. Besides, as a result of the introduction of the euro, the impact of income differentials on net capital flows has become even stronger, however, concerning the trade flows inside the euro area only. They argue this is a consequence of deeper integration of goods and financial markets, both exacerbating the trade imbalances during the catching-up period.

In fact, one of the most important benefits to be expected from constituting a monetary union is deeper goods and financial market integration (Ahearne, Schmitz and von Hagen, 2009). First, the process of financial integration reduced transaction costs on international capital flows and the introduction of the common currency further reduced the interest rate by eliminating the exchange rate risk premium. Financial integration thus substantially reduces the risk and uncertainty involved in trade transactions. The resulting decline in real interest rates together with the ongoing trend in financial liberalization have improved access to the international pool of savings, inducing poor

9_{Blanchard and Miles-Ferretti (2010) consider these types of imbalances to be “good imbalances” – a natural}

(10)

- 9 -

countries to borrow more and to maintain their bigger investment needs (Jaumotte and Sodsriwiboon, 2010; Belke and Dreger, 2011). Next to higher investments, Blanchard and Giavazzi (2002) argue that financial market integration has also led to a decrease of savings in the poorer deficit countries. To the extent that investment of these countries increase and that higher investment leads to faster expected growth, the anticipated higher permanent income boosts consumption and thus decreases their current saving.

Second, goods market integration, such as the single European market, has led to an increase in the demand elasticity of substitution between domestic and foreign goods.10_{The higher this}

elasticity, the smaller the required future price cuts in order to sell the additional output in the future and so the more attractive investment is in this period. In other words, the lower the relative price of domestic goods in the future the less attractive it is to invest in the production of domestic goods. Increased goods market integration thus reduces the adverse terms-of-trade effect that a borrowing country faces when it needs to generate a current account surplus to repay its debt.11_{To summarize,}

increased economic integration is likely to result, in the poorer countries, to both higher investments and lower savings and so to a deterioration of their external accounts (Blanchard and Giavazzi, 2002; Lane and Pels, 2011).12

In addition, Ingram (1973) argues that, in the long run, convergence of the external accounts within a monetary union will only be achieved as long as the yields of external borrowing are used for productive purposes (Lane and Pels, 2011, Giavazzi and Spaventa, 2010). This idea is based on the intertemporal budget constraint, first explained in this context by Blanchard and Giavazzi (2002). They state that a rise of external debt is only sustainable when it is accompanied by a proportional growth of national wealth. Satisfying the intertemporal budget constraint implies that today’s incurred liabilities will match the future (discounted) positive net exports. Therefore, countries must use a significant amount of the resources it borrowed to raise the potential in the production of goods that can be exported. If the country is borrowing to finance the production of non-tradable goods, the intertemporal budget constraint tends to be rejected and it might be unable to generate the required trade surpluses in the future (Giavazzi and Spaventa, 2010). As these goods are

10_{Beyond the elimination of tariffs and a stricter enforcement of competition rules across the European and}

Monetary Union, factors such as the extension of distribution networks have led to goods being closer substitutes (either in product or in geographic space), and thus to a higher elasticity of demand for each good (Blanchard and Giavazzi, 2002).

11_{An adverse terms of trade effect occurs when the value of a country’s imports of goods and services are}

greater than the value of its exports.

12_{Blanchard and Giavazzi (2002) studied the effects of European integration on the current accounts of EU}

member countries and found that the increased European integration has indeed led, in the poorer deficit countries, to both higher investments and lower private saving.

(11)

- 10 -

consumed domestically, foreign financing of their production is equivalent to borrowing money abroad to finance consumption (Belke and Dreger, 2011). 13

Not in conformity with the intertemporal approach of the current account, Giavazzi and Spaventa (2010) note that in Ireland and Spain, among others, fast economic growth was led by a construction boom that was accompanied by a large expansion of domestic credit. As the output of construction – housing – is largely a non-tradable good, Ireland and Spain provide good examples of the use of foreign capital to an extent incompatible with the intertemporal budget constraint. The case of Greece was different. With normal investment rates near the Euro area average, Greece was just not saving enough. Without public and private savings, foreign capital was used to finance Greek private consumption and budget deficits. Portugal’s imbalances are similar to those of Greece with its high consumption rate. To this extent, current account positions of the four PIGS countries were not part of the typical convergence model and have become unsustainable (Giavazzi and Spaventa, 2010).14

2.1.2 Competitiveness factors

Although many studies have often neglected the competitiveness channel as a driver of trade imbalances, Comunale and Hessel (2014) and Belke and Dreger (2010), among others, partly attribute the divergences in competitive positions to the increasing trade imbalances. According to them, the real exchange rate affects the trade position of most euro area member states – where a real depreciation coincides with an improvement of the trade balance and vice versa. Moreover, the competitiveness effect may have been intensified in the Eurozone as the introduction of the common currency reduced transaction costs and the deregulation in goods and financial markets has enforced competition (Belke and Dreger, 2011).

In fact, current account and trade imbalances coincide with changes in real exchange rates and may point to shifts in competitiveness within the Eurozone (Arghyrou and Chortareas, 2008). According to standard models of trade, real exchange rate appreciations shift demand from domestic to foreign goods. In case of a common currency, however, real exchange rate fluctuations are represented by changes in relative prices and unit labor costs. Hence, deficit countries in the Eurozone have become less competitive because domestic prices increased more than foreign prices. In other words, with the introduction of the euro, surplus countries – which kept their inflation rates low – realize a competitive advantage which leads to an improvement of their trade balances.

13_{Blanchard and Miles-Ferretti (2009) consider these types of imbalances to be an example of “bad}

imbalances”. “Bad imbalances” reflect distortions, externalities and risks at the national and international level.

(12)

- 11 -

Conversely, deficit countries, with high inflation rates, lost competitiveness and thereby increased their trade deficits (Gunnella et al., 2015).

In part, the observed divergence in price competitiveness in the Eurozone can be interpreted as sings of the catching-up process. According to the Balassa-Samuelson hypothesis, countries with higher levels of productivity growth will experience a rise of the overall price level. Higher prices subsequently induce wages to increase, causing the real exchange rate to appreciate and competitiveness to deteriorate. In view of the catching-up argument, this implies that the poorer deficit countries – who are also in a catch-up process as their rates on return are higher – rather suffer losses in price competitiveness than their richer trading partners. However, Belke and Dreger (2013) note that the deterioration of competitiveness could also be driven by other factors, such as excessive nominal wage growth or excessive fiscal spending.

In addition, Chen, Miles-Ferretti and Tressel (2013) note that the losses in price competitiveness are derived from over-optimism and excessive real exchange rate appreciation. The Eurozone system created optimistic expectations regarding the rapid convergence of the poorer countries with the richer countries. These over-optimistic expectations of faster catch-up together with the sharp drop in real interest rates generated consumption and housing booms in several counties. As economic agents were too optimistic, they anticipated growth and allowed wages and prices to rise, while productivity growth was lagging behind. The resulting real exchange rate appreciation deteriorates competitiveness and crowds out manufacturing and export activities, resulting in a deterioration of the external account. The fact that significant external borrowing only started after relatively poor countries joined the European Monetary Union (EMU) suggests that it was their EMU-membership that generated the change in expectations about future growth (Chen, Miles-Ferretti, Tressel, 2013).

2.1.3 Macroeconomic policies: fiscal imbalances

Later on, and especially after the crisis, theories started to focus more on the consequences of economic policy mistakes and “bad behavior” of deficit countries, such as excessive fiscal spending, lack of competitiveness and export capacity, excessive nominal wage growth and poor allocation of financial resources raised in financial markets (Uxó, Paúl and Febrero, 2011). Along similar lines, Eichengreen (2010) argues that the catching-up and convergence arguments seem to be a bit naïve and that rapid productivity growth in the deficit countries turned out to be a mirage. Moreover, economic convergence between the relatively poorer and richer countries is conditional not just on the gap in per capita income but also on the gap in the quality of policies and institutions. Although the latter examples of imprudent macroeconomic policies are already incorporated in – or directly related to – the catching-up and competitiveness arguments, excessive fiscal spending, or rather the

(13)

- 12 -

relative fiscal position, has not received attention yet in this study.15_{Besides, as the common}

currency reduced sovereign risk premium and substantially improved borrowing conditions, it eased the financing of fiscal deficits Canale and Marani, 2014).

The theoretical literature suggests that the potential effects of the relative fiscal position are twofold. On the one hand, the effect of fiscal policy on the trade balance can be easily obtained via the basic income identity, where the ‘twin-deficit’ hypothesis states that policies to improve the government budget position simultaneously improve the trade balance position. Fiscal expansion, for instance, boosts government spending and domestic consumption. The resulting higher demand for money increases the interest rate. High interest rates, in turn, attracts foreign capital and this capital is used to finance the excess of imports (Gruber and Kamin, 2007; Chinn and Ito, 2008). Appendix A explains the derivation of the ‘twin-deficit’ hypothesis by use of the national income identity in more detail. On the other hand, sounder macroeconomic policies such as positive fiscal budgets contribute to lower costs of foreign borrowing. The lower country risk premium should then, to the extent that lower interest rates increase investments, cause a deterioration of the trade balance (Jaumotte and Sodsriwiboon, 2010). In practice, the relative strength of both mechanisms, and thus the net impact of fiscal policy, depends on the country characteristics. Particularly in the Eurozone, however, the empirical literature provides strong evidence for the ‘twin-deficit’ hypothesis (Monacelli and Perotti, 2007; Abbas et al, 2010; Corsetti and Miller, 2006). Several studies are considered in the next subsection.

2.2 Empirical literature on trade imbalances

In addition to a detailed notion of the theories underlying the emergence of trade imbalances, it is important to be aware of the current stance of the empirical literature. The coming section will therefore give an overview of the most relevant studies related to my research methodology and consists of two parts. The first part attempts to find the relative importance of the theoretical determinants of trade imbalances described before. This is done by shortly revising some studies that empirically examined the effects of catching-up, competitiveness and fiscal imbalances on trade balance positions of euro area member countries. Moreover, the relevance of these relationships is particularly important for the empirical model that I introduce and discuss in the sections hereafter. The second part includes another strand of literature and focusses more on the presence of irrevocably fixed exchange rates, and later on the introduction of a common currency itself, rather

15_{Obviously, a poor allocation of resources raised in financial markets is related to the improper functioning of}

the caching-up and convergence mechanism, whereas a lack of competitiveness and export capacity, and excessive wage growth are related to the competitiveness component. All these types of international distortions can be seen – in the view of Blanchard and Miles-Ferretti (2009) – as types of “bad” imbalances.

(14)

- 13 -

than on the general determinants of trade imbalances. As this part concerns the studies that explore both the direct and indirect effects of a common currency on trade imbalances, it perfectly addresses my research topic and is therefore of crucial interest.

2.2.1 Catching-up, competitiveness or other drivers?

Blanchard and Giavazzi (2002) took a closer look at the current account deficits of some European countries in the 1980s and 1990s. In line with the catching-up theory, they conclude that, within Europe, capital flows follow differences in per capita income. Lane and Pels (2011) later updated and extended their empirical work by examining current account imbalances in Europe over 1995-2007, together with the underlying savings and investment rates. Their estimates show that current account imbalances indeed follow differences in relative income where lower relative income is associated with larger current account deficits. In sum, their empirical analysis affirms that the main result of Blanchard and Giavazzi (2002) has continued to hold, with capital flowing downhill from high-income countries to lower-income countries. Investments are expected to exceed savings in the latter countries, inducing external deficits in the catching up period.

Arghyrou and Chortareas (2008), on the other hand, considers the dynamics of current account adjustment and the role of competitiveness in current account determination in the Eurozone. After controlling for the effects of income growth, their empirical findings show that a negative relationship exists between the movements of the real exchange rates and the current account in the majority of the EMU-member countries. Moreover, they find that the two groups of countries with systematically improving/deteriorating current account balances during the post-EMU era period correspond to the two groups of countries that experience persistent real exchange rate deprecations/appreciations. Overall, they argue that relative competitiveness can offer further insights, beyond the effects of the income catch-up process, relevant to the trade imbalances issue observed in the euro area (Arghyrou and Chortareas, 2008).

Other studies, such as Belke and Dreger (2013), attempt to find the relative importance of catching-up and competitiveness. Their results show that both components are statistically significant with the expected signs, however, a one percentage change in competitiveness relative to euro area average has a larger effect on the current account than a one percentage change in real per capita income relative to the average – the catching up component. Moreover, the effect of competitiveness is more visible for the deficit countries, while the income effect of the catching-up argument is more important for the surplus countries (Belke and Dreger, 2013). A later study by two analysts of the Dutch Central Bank (2014) asked themselves the same question – whether the trade imbalances experienced by the Eurozone can be attributed to price competitiveness or what they call the ‘financial cycle’ – but conclude the opposite. Although differences in price competitiveness have

(15)

- 14 -

a clear influence, differences in domestic demand driven by the financial cycle are more important, thereby pointing to the catching-up argument. In this view, net capital flows, or the financial cycle, follow differences in per capita income and tend to increase domestic credit. Domestic credit booms, in turn, tend to boost domestic demand and widens external accounts. Their results, however, call for more emphasis on macroeconomic policy differences, in addition to the current attention for competitiveness and relative income levels (Comunale and Hessel, 2014).

For instance, Beetsma, Giuliodori and Klaassen (2007) investigated the consequences of fiscal policy shocks on trade balance positions in the European Union. Their findings show that a 1% increase in public spending deteriorates the trade balance by approximately 0.7%, thereby pointing to the potential relevance of the ‘twin-deficit’ hypothesis. Other studies on fiscal policy and the trade account broadly find the same positive correlation, although their findings vary in the magnitude of this effect (Monacelli and Perotti, 2007; Abbas et al, 2010; Corsetti and Miller, 2006). Moreover, Abbas et al. (2010) observed a stronger association between the current account and fiscal policy particularly for emerging economies. A possible explanation is that, in emerging and low-income economies, public spending includes a significant share of foreign goods so that fiscal expansions are more likely to spill over into imports than is the case for more advanced economies (Abbas et al, 2010).

2.2.2 Irrevocably fixed exchange rates and trade and current account adjustment

Friedman (1953) was one of the first authors that examined the relationship between exchange rate variability and external account adjustment. He claimed that a more flexible exchange rate regime allows for prompt and continuous change in relative prices and thereby promote rapid external adjustment. Despite its strong intuitive appeal, the empirical support for this idea appears to be mixed (Berger and Nitsch, 2010). Chinn and Wei (2008), for instance, examined whether the rate of current account reversion depends upon the degree of exchange rate fixity over the 1971-2005 period and found, not in conformity with Friedman’s theory, no robust monotonic relationship between exchange rate flexibility and current account adjustment. In addition, as the essay of Friedman (1953) was written in an era of limited financial integration, the results could be different in today’s world with substantially more cross border capital flows and integrated goods and financial markets (Chinn and Wei, 2008). A later study by Ghosh, Terrones and Zettelmeyer (2008), on the other hand, side with Friedman, by claiming that flexible exchange rates indeed facilitates the adjustment of external imbalances in trade and current accounts. In sum, they conclude that large current account imbalances are far less prevalent under floating regimes than they are for fixed exchange rate regimes.

(16)

- 15 -

Irrevocably fixed exchange rates, like opting for the euro, has aroused new interests in this field of research. Although the bulk of literature on the trade effects of the introduction of the euro focusses more on its effect on trade as such, there are several studies that rather attempt to capture its potential effect on the imbalances in trade. 16_{Berger and Nitsch (2010), for instance, and most in}

line with this study, studied the effect of the introduction of the euro on the dispersion of trade. They use bilateral trade data on 18 European countries from 1948 to 2008 and they observe that, as a result of the introduction of the euro, trade imbalances among the euro area members widened considerably, even after allowing for permanent asymmetries in trade competitiveness. The next subsections will discuss their methodology in more detail.

The results of Schmitz and von Hagen (2011) are in line with the results of Berger and Nitsch (2010). They use panel estimations to examine the trade balance positions of the EU-15 countries over the time period 1981-2005. Moreover, they distinguish between balances against the Eurozone and the rest of the world. In line with the catching-up theory, their results provide evidence that European trade balances follow differences in per capita income. Moreover, this effect is stronger for Eurozone members than for countries that chose to stay outside the monetary union, even before the introduction of the euro (e.g. through the European Exchange Rate Mechanism in the run-up to the monetary union). They interpret this as evidence of a deepening in financial market integration among the countries that now constitute the EMU such that the observed imbalances under member countries should be regarded, as signs of the proper functioning of the monetary union.

In like manner, Decressin and Stavrev (2009) observe that current account divergences in EMU countries have widened and that this is a trend that dates back to the early 1990’s.17_However,

a certain trend can also be observed for a broad sample of advanced countries outside the euro area and relative to this sample, current account imbalances across EMU members have not grown. Moreover, they find that current account shocks among euro area member countries are lower than for the other sample of advanced economies (Decressin and Stavrev, 2009).

Finally, Slavov (2009) examined a panel of 128 countries over the period 1976-2005 and found that common currency participants had larger current account imbalances, but only if the adopting country is situated in Europe or has a relatively high per-capita income. In addition, he found that in a monetary union, the current accounts of member countries become more sensitive to

16_{As the literature on trade as such is too broad, these studies are left out in order to stick to the imbalances}

issue.

17_{Their study proposes to investigate cross-country current account divergences and dynamics in 11 EMU}

countries (Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, and Spain) from 1970 to 2007 and compare these findings with those of 13 other advanced countries (Australia, Canada, Denmark, Iceland, Israel, Japan, Korea, New Zealand, Norway, Sweden, Switzerland, United Kingdom, and United States) with more flexible exchange rates.

(17)

- 16 -

economic fundamentals, such as disparities in relative income, relative GDP growth rates, fiscal balances and demographic structures, than non-participants (Slavov, 2009).

(18)

- 17 -

3. Data and methodology

In what follows, I add to the literature on fixed exchange rates and trade imbalances along various dimensions. First, I make use of the natural experiment provided by European monetary integration by focusing exclusively on European countries, some of which have deliberately adopted the euro as their new currency. In addition, I take a detailed look at bilateral trade relations in order to allow for a more comprehensive analysis of the effects of exchange rate conditions than the study on overall trade balances. Finally, I introduce an extended model where I combine the two strands of literature discussed before by adding potential determinants of trade imbalances to the model of Berger and Nitsch (2010) – with only a euro-dummy.

The coming section is structured as follows. The first subsection discusses the empirical models and gives a detailed description of the relevance and the shape of the included variables. The second subsection, in turn, will focus on the research methodology and will shed some light on the importance of trends and fixed effects when empirically examining trade imbalances. The subsection thereafter discusses the derived data in more detail and presents the sources of raw data. Finally, section 3.4 presents some descriptive statistics.

3.1 Models

As this study is initially interested in the effect of the introduction of the euro – or EMU membership – on bilateral trade imbalances, I first estimate variants of model (1). The regression function now is largely similar to the model of Berger and Nitsch (2010) and takes the following general form:

Model 1: �𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟� = 𝛼𝛼 + 𝛽𝛽1∗ 𝐸𝐸𝐸𝐸𝑇𝑇𝐸𝐸𝑟𝑟𝑟𝑟𝑟𝑟 {+ 𝜆𝜆𝑟𝑟 + 𝜂𝜂𝑟𝑟𝑟𝑟+ 𝜏𝜏𝑟𝑟𝑟𝑟∗ 𝑡𝑡 } + ε 𝑟𝑟𝑟𝑟𝑟𝑟

Where the dependent variable is the absolute value of the normalized bilateral trade balance and

euro is a dummy that equals 1 if both trade partners are using the euro at time t and zero otherwise.

The signs within parentheses contain a comprehensive set of fixed effects and trends, of which their relevance will be discussed in a later subsection.

In a second exercise I examine the euro-impact in more detail by adding additional country-pair specific determinants of trade imbalances to the model. The regression functions are now variants of the following general form:

(19)

- 18 -

Where Z is an additional control variable that is typically entered as the pair-wise difference in values between the reporter country r and partner country p.

Finally, by including interaction terms between the euro-dummy and some of the control variables, Z, I capture possible changes in the effects of the other explanatory variables after the introduction of the euro. Hence, by estimating variants of equation (3), I attempt to identify the mechanisms through which the euro may have contributed to the rise in trade imbalances.

Model 3: �𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟� = 𝛼𝛼 + 𝛽𝛽1 𝐸𝐸𝐸𝐸𝑇𝑇𝐸𝐸𝑟𝑟𝑟𝑟𝑟𝑟+ 𝛽𝛽2 𝑍𝑍𝑟𝑟𝑟𝑟𝑟𝑟+ 𝛽𝛽3 𝑍𝑍𝑟𝑟𝑟𝑟𝑟𝑟∗ 𝐸𝐸𝐸𝐸𝑇𝑇𝐸𝐸𝑟𝑟𝑟𝑟𝑟𝑟

{+ 𝜆𝜆𝑟𝑟 + 𝜂𝜂𝑟𝑟𝑟𝑟 } + ε 𝑟𝑟𝑟𝑟𝑟𝑟 ,

3.1.1 Dependent variable: the bilateral trade balance

In line with Berger and Nitsch (2010), the dependent variable is the bilateral trade balance between a reporter country r and a partner country p, defined as the difference between country r’s exports to

p and country r’s imports from p in a given year t. To account for differences in the importance of

trade relationships across partners, the trade balance is normalized by the total value of bilateral trade between country r and p. This is the natural choice as larger values of trade balances often coincide with larger bilateral trade volumes between two countries. The bilateral trade balance is formulated as follows;

| 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 | = |_ExportsExports_rptrpt_{+ Imports}− Importsrpt_rpt |

As my research question already suggest, I expect the introduction of the euro to have a diverging effect on member counties’ trade balances. This implies that the introduction of the euro should have induced some countries to end up with larger trade deficits, while other countries simultaneously end up with larger trade surpluses. As a matter of course, when using bilateral trade balances, a larger bilateral surplus for the reporting country implies a larger bilateral deficit for the partner country and vice versa. In order to capture this diverging effect for both countries simultaneously, absolute values are employed where larger magnitudes of the variable indicate greater imbalances in bilateral trade (Berger and Nitsch, 2010). Moreover, when talking about imbalances, the only important aspect is the magnitude of the bilateral trade balance and not its direction (Berger and Nitsch, 2010). In order to reinforce the choice for absolute values rather than nominal values, I perform some simple regressions in appendix B.

(20)

- 19 -

3.1.2 Variable of interest: the euro-dummy

The inclusion of the Euro-dummy is more straightforward. As this study attempts to find an answer on the question whether the introduction of the euro has contributed to larger trade imbalances within the Eurozone, the coefficient on this variable is of crucial interest. Euro is a dummy that equals one if both countries adopt the euro in time t and zero otherwise. As I justified the use of absolute values in appendix B, a positive coefficient now indicates either an increase of a bilateral trade surplus or an increase of a bilateral trade deficit, where both developments simultaneously imply an increase in trade imbalances since the introduction of the euro.

3.1.3 Control variables:

Although I fist replicate and amend the standard model of Berger and Nitsch (2010), by use of model (1), I think their model is thin and incomplete. They just added one independent variable, the euro-dummy, while I think there are more relevant variables determining trade imbalances – and those in the Eurozone in particular. Hence, by including more pair-specific determinants in models (2) and (3), I expect to find a more precise impact of the introduction of the euro on trade imbalances than was found by Berger and Nitsch (2010).

The theoretical part in this study, section 2.1, suggests that trade imbalances in the Eurozone are predominantly driven by two components; catching-up and competitiveness. Therefore, variables measuring relative income and competitiveness will be included in the model. Furthermore, the impact of differences in macroeconomic policies, and then especially the relative fiscal position, has proved to be an important indicator (Monacelli and Perotti, 2007; Abbas et al, 2010; Corsetti and Miller, 2006). However, since fiscal imbalances might be considered endogenous relative to trade imbalances (e.g. because governments might pursue a current account target for fiscal policy), the results should be interpreted with cautious (Schmitz and von Hagen, 2010). As the inclusion of the relative fiscal position is furthermore not of primary interest for this study – on the EMU-effects in particular – it is merely included as a control variable and will not be interacted with the euro-dummy in model (3).18

Because the study takes bilateral trade imbalances as the indicator of trade imbalances, all control variables are entered as the pair-wise difference in values between the two trading partners. Again, as I am only interested in the magnitude of these differences, absolute values are employed.

18_{Catching-up and competitiveness, on the other hand, are particularly important for the Eurozone (in the}

(21)

- 20 -

3.1.3.1 Relative income: GDP per capita

According to the literature, the rise in trade imbalances in the Eurozone follow differences in per capita income. In fact, economic theory suggests that differences in per capita income reflect differences in development stages and the associated growth prospects. Relatively poorer countries are the countries where the rates of return are higher. These countries are therefore rather willing to run larger trade deficits in order to achieve these higher growth rates in the future. This is known as the catching up component and should be, in part, seen as part of an healthy convergence process between economies (Lane and Pels, 2011; Blanchard and Giavazzi, 2002).

The analyses of the catching up component refers in most studies to relative income, where relative income is measured in terms of differences in GDP per capita between a reporter country r and a partner country p: log�𝐺𝐺𝐺𝐺𝑃𝑃𝑟𝑟− 𝐺𝐺𝐺𝐺𝑃𝑃𝑟𝑟� (Jaumotte and Sodsriwiboon, 2010; Belke and Dreger,

2011; Nieminen, 2014).19_{Since economic agents are furthermore not able to instantly react on}

changes in growth rates and the associated rates of return, lagged variables are included as well. Moreover, the catching up period accompanied by higher external deficits often lasts longer than 1 year as countries do not immediately achieve the intended higher growth (d'Arvisenet, 2008). Overall, I expect larger income differentials to be associated with larger trade imbalances;

β = Δ �𝑇𝑇𝑟𝑟𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟� Δ log �𝐺𝐺𝐺𝐺𝑃𝑃𝑟𝑟−𝐺𝐺𝐺𝐺𝑃𝑃𝑟𝑟� > 0 Δ �𝑇𝑇𝑟𝑟𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟� Δ log �𝐺𝐺𝐺𝐺𝑃𝑃𝑟𝑟−𝐺𝐺𝐺𝐺𝑃𝑃𝑟𝑟� 𝐿𝐿1 > 0 Δ �𝑇𝑇𝑟𝑟𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟� Δ log �𝐺𝐺𝐺𝐺𝑃𝑃𝑟𝑟−𝐺𝐺𝐺𝐺𝑃𝑃𝑟𝑟� 𝐿𝐿2 > 0

Where L1 and L2 refer to lags 1 and 2 respectively.

3.1.3.2 Competitiveness: Unit labor costs

Differences in competitiveness pose another important indicator of trade imbalances within the Eurozone (Belke and Dreger, 2013; Comunale and Hessel, 2014). Especially in the case of a common currency, where nominal exchange rates cannot fluctuate in order to restore competitiveness, productivity becomes an important factor. The best way to account for that in the model is by including a variable that measures the relative competitiveness between two trading partners. Unit labor costs can be considered a good indicator describing competitiveness and they are calculated as the ratio of labor costs to labor productivity (Merstina and Jänes, 2012, Felipe and Kumar, 2011; Irvin, 2011).

19_{Using log instead of the nominal value is because of interpretative reasons. Since GDP per capita is measured}

in terms of US dollars, a one dollar change would not make a big difference, what makes it more favorable to measure this effect in terms of percentage changes. A 1% change in income differentials is now associated with a 0.01*β change in the dependent variable.

(22)

- 21 - 𝑁𝑁𝐸𝐸𝑁𝑁𝑁𝑁𝑇𝑇𝑇𝑇𝑇𝑇 𝑈𝑈𝑇𝑇𝑁𝑁𝑡𝑡 𝐿𝐿𝑇𝑇𝐿𝐿𝐸𝐸𝑇𝑇 𝐶𝐶𝐸𝐸𝐶𝐶𝑡𝑡𝐶𝐶 = Employee′_{s compensations} Number of employees � Real GDP

Employed persons total �

In order to maintain a high level of competitiveness, labor costs should not increase faster than labor productivity on a permanent basis. Hence, countries with relatively low unit labor costs – compared to their trading partners – are generally regarded as more competitive. In order to eliminate periodic fluctuations, adjusted unit labor costs are usually analyzed (Mertsina and Jänes, 2012, Nieminen, 2014).

Although there is no consensus in the literature about the preferable form of unit labor costs, I follow Nieminen (2014) by adopting real unit labor costs. While his study also lacks well substantiated arguments for his choice, he argues that real unit labor seems to capture the aspects of competitiveness that are important for intra-Eurozone trade balances better than nominal unit labor costs. Relative competitiveness between the reporter country r and partner country p is therefore considered in terms of absolute differences in RULC; log|𝑅𝑅𝑈𝑈𝐿𝐿𝐶𝐶𝑟𝑟− 𝑅𝑅𝑈𝑈𝐿𝐿𝐶𝐶𝑟𝑟|.20 I expect

larger differences in competitiveness to positively affect trade imbalances.

β = Δ �𝑇𝑇𝑟𝑟𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟� Δ �log (𝑅𝑅𝑅𝑅𝐿𝐿𝐶𝐶𝑟𝑟−𝑅𝑅𝑅𝑅𝐿𝐿𝐶𝐶𝑟𝑟)� > 0

3.1.3.3 Macroeconomic management: fiscal imbalances

Finally, I want to control for the effect of differences in the soundness of macroeconomic management between countries. Since the emergence of asymmetries in trade could not be offset by standard policy instruments within a monetary union, fiscal policy is the most important policy tool (Jaumotte and Sodsriwiboon, 2010; ECB, 2013).21

Like to the rest of the model, also the relative fiscal position is included in absolute differences; |𝐺𝐺𝑇𝑇𝑟𝑟𝑟𝑟− 𝐺𝐺𝑇𝑇𝑟𝑟𝑟𝑟|, where the government budget balance (GB) of a respective country r or p is calculated as the difference in government’s expenditures and revenues (both as a percentage of

GDP). In this regard, larger values of the variable can be seen as larger fiscal imbalances between two trading partners. Following the ‘twin-deficit hypothesis’, I expect larger cross-country differences

20_{The use of log is explained similar to that of taking the logarithm of relative income.}

21_{Clearly, the common interest rate is set by the ECB and also exchange rate devaluation is since the}

introduction of the euro no option anymore. Hence, fiscal policy remains the most important macroeconomic policy instrument, even though fiscal policy is also very constrained by the well-known institutional barriers such as the Stability and Growth Pact (ECB, 2013).

(23)

- 22 -

in terms of government budgets – larger fiscal imbalances – to be associated with larger trade imbalances.

β = Δ �𝑇𝑇𝑟𝑟𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇_{Δ �𝐺𝐺𝑇𝑇𝑟𝑟−𝐺𝐺𝑇𝑇𝑟𝑟�}𝑟𝑟𝑟𝑟𝑟𝑟� > 0

3.2 Method

I follow the literature by applying Ordinary Least Squares (OLS) as the estimation strategy in performing the regressions (Berger and Nitsch, 2010; Schmitz and von Hagen, 2011). In addition, a comprehensive set of fixed effects will be considered, of which I discuss their relevance hereafter.

3.2.1 Fixed effects and trends

An important issue related to regression models and to panel data models in particular is omitted variables bias. In statistics, omitted variables bias occurs when a model is created and incorrectly leaves out one or more important causal factors. The ‘bias’ is then created when the model compensates for the missing factors by under- or overestimating the effect of one of the other explanatory variables. The threat within my model is that omitted upward trending variables in combination with a euro dummy – which is only one at the end of the sample – may lead to an upward bias in the estimated euro-effect. Without correcting for this, the euro dummy picks up the increasing trends in trade imbalances that are actually caused by omitted variables. The euro effect will then capture, among others, general economic integration among the member states and not merely the common currency effect (Bun and Klaassen, 2006; Gunnella et al., 2015).

Several studies tried to reduce the abovementioned endogenous effect of currency unions by including country-pair and year fixed effects (Berger and Nitsch, 2010; Bun and Klaassen, 2006; Micco et al., 2003). Controlling for country-pair specific fixed effects ( 𝜂𝜂𝑟𝑟𝑟𝑟 ) allows pair-wise imbalances to

consistently deviate from the sample average. In fact, each country-pair has its own characteristics that may influence the estimates of the variables. As some bilateral trade relationships could be structurally characterized by one directional-trade flows, the inclusion of pair-wise fixed effects severely limits the possible effect of these outliers on the results. The inclusion of time fixed effects

( 𝜆𝜆𝑟𝑟 ) controls for joint variations in trade imbalances over time and captures important determinants

such as the state of the world economy. For instance, it is likely that events such as the credit crisis of 2008 have an impact on all bilateral trade relationships, such that the inclusion of time fixed effects takes away these effects. Time- and pair-wise fixed effects are quite standard elements in panel data

(24)

- 23 -

models and will therefore be considered in order to ensure that the results can be easily compared to the existing literature (Bun and Klaassen, 2006; Micco et al. 2003; Berger and Nitsch, 2010).

In addition to the conventional fixed effect estimation I investigate another particularly important time-series characteristics, namely trends. Although the inclusion of common time fixed effects also allow for some trend T, this trend is restricted to be common to all country-pairs.22_As

omitted variables still clearly reflect time factors that vary across country-pairs this could still lead to a biased euro estimate (Bun and Klaassen, 2006; Baldwin, 2006). To deal with the heterogeneous effect of these time varying omitted components, I follow Bun and Klaassen (2006) by including an pair-specific time trend t, where the country-pair specific dependence on the trend is represented by 𝜏𝜏𝑟𝑟𝑟𝑟. Transport costs, for instance, depend on distance and the goods composition of trade, which are

both different across country-pairs. Because transport costs have decreased over time this can be considered a source of a country-pair specific trend that needs to be controlled for. Moreover, as economic integration has existed over the whole sample and gradually deepened across the country-pairs, its effect on trade imbalances may be captured more by ( 𝜏𝜏𝑟𝑟𝑟𝑟 * t ), than by the euro-dummy

𝐸𝐸𝐸𝐸𝑇𝑇𝐸𝐸𝑟𝑟𝑟𝑟𝑟𝑟.

In models (2) and (3), however, where I’m particularly interested in the impact of other relevant determinants, I only account for pair-wise and common time fixed effects. Omitted variables bias is now already mitigated as the model includes other trending variables, such as the trend in diverging levels of income and competitiveness, to explain the trend in bilateral trade imbalances. Moreover, as I include common time effects I still correct for any residual trend T common to all bilateral trade balances (Bun and Klaassen, 2006). Besides, including too many dummies – that is necessary for the inclusion of pair-specific trends – absorb too much of the useful variation of the data and therefore severely limits the effects of the other variables. Standard estimations techniques such as common time and pair-wise fixed effects do not reveal these problems (Hornok, 2011).

22_{To emphasize the importance of trends in issues like the trade imbalances in the Eurozone, I also estimate}

the consequences of including common linear time trends (T) for the euro-effect in model 1 – before I replace them by the more compassing common time fixed effects ( 𝜆𝜆𝑟𝑟 ).

(25)

- 24 -

3.3 Data

My sample consists of a homogenous set of fifteen European countries; thirteen countries (excluding Belgium and Luxembourg) which were member of the EU at the time of the introduction of the euro (9 of which adopted the Euro from the beginning, followed by Greece in 2001) plus Norway and Switzerland.23_{Restricting the sample to this group of countries has the advantage that all countries}

either share the institutional framework of the European Union or are closely associated with it (Berger and Nitsch, 2010, 2008).

The dataset covers the time period from 1980 till 2011, so that there are T=32 periods. The panel is balanced and this results, given that there are 105 country pairs, in N=3360 observations. Furthermore, as already mentioned in the introduction, trade balances of Eurozone members seem to have been converged since the start of the latest crisis around 2008 (Sinn and Valentinyi, 2013). In order to control for this occurrence, I perform regressions with and without the crisis years. Estimations without the crisis years will yield N=3045 observations.

Data for bilateral trade comes from the International Monetary Fund’s Direction of Trade Statistics (IMF DOTS), from where I obtained nominal values of bilateral exports and imports on an annual basis. Data on GDP per capita is obtained from the World Bank (2014) and is measured in terms of current US dollars. Real Unit Labor Costs were extracted from the AMECO database of the European Commission (2014). Finally, data on the countries’ government budget positions were acquired from the IMF World Economic Outlook database (WEO, 2014). Table 9 in appendix B summarizes the data as described above. In addition, consider the included data file (Excel) for a more accurate description of how the data is derived.

3.4 Descriptive statistics

Descriptive statistics for the sample are provided by table 1. The average magnitude of the bilateral trade balance is 0.24 and implies an absolute bilateral trade balance of 24% of total bilateral trade. The minimum and maximum values represent the lowest and highest values of the sample respectively. Hence, the lowest value of the trade balance is approximately 0.00005, what implies that exports and imports between trading partners are near to equal, whereas a maximum value of about 0.935 suggests almost one-way trade.24_{The descriptive statistics for the additional control}

variables relative income, competitiveness and the relative fiscal position are given in absolute

23_{The dataset of Berger and Nitsch (2010) consists of 18 European countries. Unfortunately, the only way this}

dataset was accessible for me, is by omitting Belgium, Luxembourg and Iceland. However, as these countries are relatively small, I expect their relative importance for the overall European trade imbalances issue, in real terms, to be mitigated.

24_{Lowest bilateral trade balance is the trade balance between the UK and Spain in 1985, whereas the highest}

(26)

- 25 -

wise differences in terms of GDP per capita, real unit labor costs and the fiscal budgets respectively. The mean value of the relative fiscal position, for instance, implies that the fiscal budgets of two trading partners differ, on average, by about 4.7 percentage points.

Table 1. Descriptive statistics

Variable Obs. Mean Std. deviation Min Max

Trade Balance Euro

Relative income Real unit labor costs Relative fiscal position

3360 3360 3360 3206 2735 .2407855 .16875 9975.55 5.349059 4.685869 .1931459 .3745867 9408.727 4.029933 4.632445 .0000484 0 3.250781 .0036887 0.004 .9353131 1 76558.59 23.02005 41.639 In addition, table 2 gives the correlation matrix of all the variables. It is intended to give an indication of the direction and the strength of the linear relationships between the variables. According to the matrix, all variables are positively correlated with higher values of the bilateral trade balance. This is in line with my expectations that periods typified by large mutual differences in levels of income, competitiveness and fiscal budgets are often associated with periods of trade imbalances (Eichengreen, 2010). See figure 4 in Appendix B for a more accurate view on the data. It illustrates tables 1 and 2 by graphing the evolution of the variables, represented by sample means, over time.

Table 2. Correlation matrix

Trade balance Euro Relative

income Real unit labor costs Relative fiscal position Trade Balance Euro Relative income Real unit labor costs Relative fiscal position

1 0.1348 0.3045 0.1400 0.2779 - 1 -0.0763 -0.2077 -0.0990 - - 1 0.2233 0.4821 - - - 1 0.1594 - - - - 1

Finally, to analyze the data on bilateral trade in more detail, figure 3 shows the average magnitude of bilateral trade balances distinguished by type of country-pair. Specifically, I distinguish between trade relationships for country-pairs that both adopt the euro (intra-EMU trade) and country-pairs for which nominal exchange rates have remained flexible (i.e., trade between EMU countries and non-EMU countries as well as trade between non-EMU countries).25_{Although trade imbalances}

among country-pairs within the Eurozone seem to be higher and more volatile over the full sample

(27)

- 26 - 0,1 0,15 0,2 0,25 0,3 0,35 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 EMU vs EMU EMU vs non-EMU Non-EMU vs non-EMU

period, all types of country pairs follow a similar pattern; bilateral trade imbalances commonly increased in the run-up to the introduction of the euro and continued to do so afterwards, although with lesser extent. Again, trade balance adjustments of deficit countries in the run-up to the financial crises explain the common decline in trade imbalances since the late 00’s (Sinn and Valentinyi, 2013). Hence, on the basis of figure 3 it’s hard to identify a certain common currency effect. In the next section I attempt to capture this effect empirically according to the models derived before.

Figure 3. Bilateral trade imbalances by group of country pairs

(28)

- 27 -

4. Results and discussion

The coming section presents the estimation results and consists of three parts. In the first subsection I comprehensively examine the direct effects of the introduction of the euro on trade imbalances using the basic model (1). In the second subsection, I discuss the results of the augmented model (2) and examine country-pair specific determinants in more detail. Finally, I discuss the results of the interacted model (3) in an attempt to identify the mechanisms through which the euro could have contributed to larger trade imbalances.

4.1 Trade imbalances and the introduction of the Euro

Table 3 presents the estimation results of variations of regression (1), where I directly examine the impact of the euro on bilateral trade imbalances, while controlling for several combinations of fixed effects and trends.

(1) �𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟� = 𝛼𝛼 + 𝛽𝛽1∗ 𝐸𝐸𝐸𝐸𝑇𝑇𝐸𝐸𝑟𝑟𝑟𝑟𝑟𝑟+ 𝛽𝛽2∗ 𝑇𝑇 +𝜆𝜆𝑡𝑡 + 𝜂𝜂𝑟𝑟𝑟𝑟+ 𝜏𝜏𝑇𝑇𝑟𝑟∗ 𝑡𝑡+ ε 𝑟𝑟𝑟𝑟𝑟𝑟

I start with the most parsimonious specification of equation (1); a regression of the absolute value of the bilateral trade balance on a euro-dummy and a set of pair-wise fixed effects. As shown in the first column the estimated coefficient on the euro-dummy is positive and significantly different from zero at the 1 percent level. The point estimate of about 0.06 implies that trade imbalances among euro area members are on average about 6% larger than for the rest of the sample, which appears to be quite large compared to a sample mean of 24%. In the next column, I add a linear time trend T to the model that captures any omitted upward trending variable that is common to all country pairs. The result significantly supports the trend relevance with respect to trade imbalances and implies that, on average, bilateral trade imbalances among a given country-pair increase by 0.17 percentage points per year. Although the coefficient on the euro-dummy remains significantly positive, its impact dropped half in magnitude as column (2) now displays a euro-impact of only 3.25%. In column (3) I replace the common linear time trends by common time fixed effects, which results in an even lower and insignificant euro-estimate.26

26_{Only using a linear time trend constraints the time effect to lie on a straight line, whereas estimating with}

common time fixed effects allows the coefficient pattern to be whatever the data chooses. Hence, common time fixed effects not only capture the common linear time trend but also controls for other joint variations over time.

The introduction of the Euro and trade imbalances in the Eurozone