OUP user menu

Life on the outside: economic conditions and prospects outside euroland

David Barr, Francis Breedon, David Miles
DOI: http://dx.doi.org/10.1111/1468-0327.00116_1 573-613 First published online: 1 October 2003

Abstract

Life on the outside

The European economic and monetary union (EMU) is now over 4 years old. In this paper we assess whether monetary union has begun to have significant economic effects by comparing countries in EMU with the EU countries outside. We focus principally on trade creation between EMU member countries, using a methodology that controls for the fact that the decision to join the monetary union was not random but was more likely to be taken by countries whose prospects of trading with other EMU members were already high. We find that the trade effects of monetary union are significant. We estimate that had the UK been inside EMU the sum of its imports and exports could have been substantially greater. For comparative purposes, we also make preliminary estimates of the effect of monetary union on three other dimensions of economic performance: foreign direct investment, the development of financial markets and overall macroeconomic performance, though we recognize that our ability to control for other factors is more limited in respect of these other indicators. The evidence suggests that inward investment in the countries outside would have been greater had they joined EMU, but that the impact of this on GDP would be no more than 0.3% of GDP per annum for the UK and less than that for the other ‘outs’. Financial market activity shows no clear sign of having been affected by EMU, and London’s position as Europe’s financial centre remains, as yet, largely unchallenged. On standard measures of aggregate performance – inflation, unemployment and output – no clear pattern of EMU effects has yet emerged.

— David Barr, Francis Breedon and David Miles

1. INTRODUCTION

What can we learn about the potential benefits and costs of the European monetary union by comparing the performance of the ‘ins’ and ‘outs’ over the past few years? Those who are strongly in favour of ‘outs’ staying out or of ‘outs’ jumping in implicitly take the view that the answer to this question is ‘a lot’. Those sceptical about whether the UK, Sweden and Denmark should adopt the euro point out that unemployment in those countries has stayed lower than in the euro area, that inflation has remained low and that there are no signs from financial markets that the credibility of monetary policy in controlling inflation has been harmed. The euro enthusiasts – particularly in the UK – have focused on the (largely) unwelcome strength of the ‘out’ currencies against the euro for much of the period since 1999 and have highlighted the relative weakness of foreign direct investment inflows into the ‘outs’.1 Sorting out what we can, and particularly what we cannot, learn about the implications of adopting the euro from recent economic events within Europe is important. It is what we try to do in this paper. We take the view that overall real macro performance – output and unemployment – is as yet unlikely to have been affected by the introduction of the euro in ways which make it clear how the outs and ins have done relative to each other by dint of their decisions on the currency. But in three areas – trade, foreign direct investment (FDI) flows and financial market activity – effects are potentially important and likely to show up more quickly than with the usual economic aggregates where trends and cycles having little to do with EMU probably still dominate recent outcomes. In the event it is only the trade effect that shows up clearly in the data so far: our estimates of the effect on FDI and financial markets serve mainly to indicate that, to the extent that any effects can be discerned at all, they are of a much lower order of magnitude than the trade effects. We focus mainly on the relative performance of the euro countries with the EU ‘outs’– Denmark, the UK and Sweden. But we also sometimes consider the performance of the EFTA countries.

It is important to work out at the outset what made some countries adopt the euro. Unless we have some idea about that it is hard to know how much of the recent economic performance of the ‘ins’ is a result of adopting the euro and how much is just a reflection of the forces, whatever they might be, that made countries wish to adopt the currency in the first place. So the starting place for our analysis is a consideration of what made countries ‘ins’ or ‘outs’ in the first (or in the case of Greece, second) place. In Section 2 we start the paper by looking at what distinguished ins and outs before monetary union got going. We then look at three areas where the potential impact of either adopting the euro or retaining the national currency may be large: trade, FDI and the scale and location of financial market activity. In Section 3 we see what we might learn about the impact of the single currency on bilateral trade flows. In Section 4 we concentrate on FDI flows. In Section 5 we briefly survey the evidence on financial market location and activity, while Section 6 widens the analysis to look at overall economic performance in the ins and the outs.

In Section 7 we draw conclusions, paying as much attention to what recent trends do not tell us as to the more positive messages from the data.

2. WHY IN? WHY OUT?

Before we can undertake an analysis of the relative performance of EMU ins and outs, we need to have some explanation of why the outs chose to be outs and the ins chose to be ins. Without such an explanation, it is difficult to argue that any observed differences between the ins and outs is due to monetary union as opposed to another factor that explains both the observed difference and the propensity to join EMU.

2.1. An analytical framework

Our starting point for such an explanation is the work of Alesina and Barro (2000) (which is itself an extension of the work of Mundell, 1961). They identify the main economic determinants of the propensity to join a currency union as:

  1. Trade – the more two countries trade already, the greater will be the benefit of forming a currency union (though this result depends on ‘reasonable’ assumptions about the elasticities of substitution). Alesina and Barro also argue that other variables that may influence trade (such as distance between countries) could have an independent effect on the desire to join a currency union – we ignore those effects in this section.

  2. Co-movements of output – the greater the co-movement of output between countries, the smaller will be the cost of giving up monetary independence through a currency union.

  3. Co-movements of prices – as with output co-movement, greater price co-movement reduces the cost of giving up an independent monetary policy.

  4. History of high and volatile inflation – a country may wish to join an existing currency union, or form a union with a larger partner, in order to gain the discipline and credibility of the anchor currency's monetary policy.

For our purposes, the first three of these determinants are the most relevant as EMU, arguably, has no one ‘anchor’ currency to which other countries are linked. (Although Germany was the largest single economy in the union, the fact that a new currency and central bank were created suggests that EMU cannot be simply seen as a DM currency zone). Table 1 outlines estimates of these three determinants for countries within the EU and EFTA (excluding Luxembourg and Liechtenstein for data reasons).2

View this table:
Table 1. Indicators of the propensity to form a currency union (average for the period 1978–91)

Looking at Table 1 in detail, in the row labelled ‘All’ we present the average value of each of the three determinants (trade, output co-movement and price co-movement) for all possible bilateral pairings in our sample. The row labelled ‘EMU’ then shows the same calculation but now only for bilateral pairings involving two EMU countries. Focusing on the non-EMU members of the EU we also show the average value of each determinant for all pairings involving each of these countries with every EMU member. So ‘UK with EMU’ shows the average value of our determinants for all pairings involving the UK and an EMU country – similarly for Sweden and Denmark. For example, for trade we calculate bilateral trade as a share of GDP for every possible pairing in our sample (272 combinations) over the period 1978 to 1991 (up to the signing of the Maastricht Treaty). The row labelled ‘All’ shows the average of all these combinations, the row labelled ‘EMU’ shows the average bilateral trade only for combinations involving two (future) EMU members. The ‘UK with EMU’ row shows UK average trade share with each EMU country, similarly for Sweden and Denmark.

Additionally we test to see if the EMU-ins are significantly different from the outs using a Z-test. For each indicator this involves finding the average value of the indicator for bilateral pairings involving two EMU-ins and comparing that with the average value of the indicator for all bilateral pairings involving at least one EMU out. The individual country tests (for UK, Sweden and Denmark) simply compare the EMU-in sample with the average indicator value for every pairing involving each of those countries. The Z-test is then a test of whether the mean of the EMU-in sample is significantly different from the mean of the out, or individual country, sample.

We find that both in terms of trade and output co-movement the countries that formed EMU were better placed to do so than the rest of our sample (on average). In particular, there is a highly significant relationship between output co-movement and subsequent EMU membership. This perhaps indicates that the issue of economic convergence had a significant impact on the decision to join EMU. Price co-movement (arguably the least important of the three factors, not least because of its interaction with the credibility effect), on the other hand, appears to have no significant link with EMU membership.

A potential problem with our results is that while they appear to help explain whowill be in the currency union they do not tell us why the union was formed when it was. Given our use of time series averages to construct our indicators, the timing of EMU is difficult to gauge. However, Table 2 indicates that when we compare the period 1978–88 with the period 1988–98, all of our indicators are more conducive to a currency union in the later period, indicating perhaps that costs of monetary union have fallen and the benefits risen over time. This goes some way to explaining why monetary union did not occur earlier.

View this table:
Table 2 Indicators of the propensity to form a currency union (period averages, all countries)

Overall, it seems that we can find economic variables that help explain why some countries chose to join EMU and some did not. In particular, output co-movement proves to be an important predictor of the subsequent decision to join. Not only does such a result undermine the argument that the decision was purely political and unrelated to economic considerations, these predictors are invaluable in determining whether EMU has had significant economic effects. In essence, they help us unravel the economic impact of EMU from the economic determinants of EMU. We discuss this in more detail below.

3. DID BEING OUT OF EMU INFLUENCE TRADE?

One of the key benefits of joining a currency union is its potential effect on trade. A large and growing literature has identified the existence of different currencies as a surprisingly large barrier to trade. If that is so, trade by the EMU outs might be expected to suffer relative to the ins.

Rose (2000) was the first to estimate the effect of currency unions on trade and his work found a remarkably large impact (over 200% increase in trade for members of a union). His approach, based on a standard gravity model (where bilateral trade flows are positively related to GDP of the two countries and negatively related to their distance from one another), attempts to control for the main factors that influence trade, and then includes a dummy variable for currency union. The coefficient on this dummy is then interpreted as the currency union effect on trade. Following Rose's work a number of authors have attempted to overturn his result with varying degrees of success. But most studies have actually confirmed a large currency union effect. However, two important criticisms of Rose's study have arisen. The first concerns sample selection: Rose's sample of currency unions was small (less than 1% of the total sample) and dominated by developing countries. Thom and Walsh (2001) studied Ireland's break from sterling and found no significant trade effects, suggesting perhaps that Rose's estimates do not apply to developed country unions. The second criticism centres on the endogeneity issue discussed below. If it were the case that countries expecting a large increase in trade tend to form currency unions, then the estimated relationship between trade and currency unions cannot be interpreted as a currency union effect.

Ever since Rose's results first appeared, the most significant criticism of this approach has been the issue of endogeneity. If this is a problem, the estimated impact of currency unions may to a significant extent be due to the impact of a third variable that predicts entry into a currency union and more trade. To assess the impact of EMU on trade we need to address this endogeneity issue. While this problem is familiar to most economists, in the context of a panel estimate (which combines cross section and times series) it becomes slightly more involved.

The underlying problem can be expressed most simply using two equations:

TRADE =α+β(EMU) +γX+θZ+ɛ1

EMU =ψ+δZ+χW+ξ2

Here trade is a function of EMU membership, variable X and variable Z (Equation 1). EMU entry is a function of variable W, and is also a function of variable Z (Equation 2). When we estimate the impact of EMU on trade we must attempt to control for all possible variables that can influence trade other than EMU. In particular, were we to leave out variable Z from our model, then the estimate of β (the trade impact of EMU) would be biased since it would include both the impact of EMU and, indirectly, the impact of variable Z. The problem we face is that unless we control for all the variables that influence trade, we cannot be sure that our estimated EMU effect is correct since we may have excluded a variable like Z. The first solution, that of Rose (2000), is to include an extensive range of variables that might influence trade. This approach was criticized by Persson (2001) who pointed out that even if the list of variables was exhaustive, if some of those variables influenced trade in a non-linear way, but were included as having a linear relationship, then the estimate of the currency union effect could still be biased. An alternative approach is to include fixed effects in our model that control for any differences in characteristics of EMU and non-EMU countries. This involves allowing for a dummy variable for every country and means that the EMU effect is only significant if there is a change in trade post-EMU rather than just higher trade on average for EMU-ins. As a result, even if we cannot explicitly identify all the differences in characteristics between countries, the fixed effect will pick them up.

However, even allowing for fixed effects there is still a potential endogeneity problem. This arises if the variable we inadvertently exclude predicts both the decision to join EMU and higher trade post-EMU. For example, if a country observes a change in its circumstances that will increase trade with EMU countries in the future, it may choose to join EMU as a way of maximizing the benefit of that change. As a result, we will observe the decision to join EMU being followed by higher trade, so even the fixed effect estimate will ascribe the increase in trade to EMU. To get around this problem, we need an instrumental variable – something that predicts EMU entry, but cannot have been influenced by the potential trade increase post-EMU (such as variable W in Equation 2). If we can find such a variable (or variables), then we can use it to assess the probability that a country will join EMU, and see if that probability predicts higher trade post-EMU. If it does then we can ascribe the increase in trade to EMU membership since, if the instrument is valid, it could not have been influenced by the post-EMU increase in trade. It is this instrumental variables approach that we take in this paper.

3.1. The trade impact of EMU

In this and the following section we aim to estimate the effect of EMU on trade. Our approach is to start by estimating the standard gravity model of Rose (2000) for the period 1978Q1 to 2002Q1 for a panel of European countries (quarterly data is used to maximize our post-EMU sample and extend it to the introduction of euro notes and coins). Our sample of countries is once again all of the EU and EFTA except Luxembourg and Liechtenstein, thus giving us 11 EMU ins and 6 outs. Given that we are looking at every bilateral combination of our 17 countries, we have 136 different time series each covering the period 1978Q1 to 2002Q1 giving us a total panel sample of 13,192 observations.

Since the range of variables that might influence trade is myriad, the range of possible specifications of the gravity model is virtually limitless. Therefore we narrow the field by focusing on Rose's original simple specification (though we exclude former colonizer effects for obvious reasons). The column headed OLS in Table 3 outlines the results.3

View this table:
Table 3 Gravity trade model (log of bilateral trade as a function of)

Generally speaking, the model has similar coefficients to Rose's original model with all the variables entering significantly and with the expected sign. We also find a positive and significant effect for membership of the monetary union of about 29%4 (and of the EU). An alternative specification including fixed effects (which requires us to drop variables such as distance that have no time series variation) gave an EMU effect of 27%.

Our estimates of the currency union effect are significantly smaller than that found by Rose (29% rather than 240%), though offset somewhat by the fact that the exchange rate volatility effect is larger in the Rose specification (12% trade impact for every 1% change in standard deviation rather than the 2% trade impact found by Rose). Our high estimate for the impact of exchange rate volatility is a little surprising given that De Grauwe and Skudelny (2000) find a much smaller (but still significant) impact of exchange rate volatility using a similar sample to ours. Certainly, in our robustness checks using alternative specifications, this volatility effect was smaller (though still significant) in most of the other specifications we looked at (for example, as low as 0.01 in the fixed effect version). These checks also confirm that, unsurprisingly, the volatility effect interacts quite strongly with the EMU effect (excluding volatility from the Rose specification increases the EMU effect to about 40%). A number of these checks are presented in the appendix.

3.2. Instrumental variables estimation

While our version of the gravity equation avoids one of the criticisms of this type of currency union model – namely an excessive reliance on currency unions involving developing countries – it does not, as it stands, avoid the second: we have not yet allowed for the possibility that the propensity to join the currency union is related to factors that increase trade.

As discussed above, instrumental variables can help to overcome this problem and so we use the price and output co-movement variables discussed in Section 2 as instruments.5 As we have seen, output co-movement in particular has a very strong relationship with subsequent EMU entry, so it and price co-movement seem good candidates as instruments. Certainly, by focusing on co-movement for the period 1978 to 1991 (before the Maastricht Treaty was signed) we can be fairly confident that these co-movement variables are independent of the prospective trade impact of EMU. So, even though there is evidence that increased trade can increase co-movement (see, e.g., Frankel and Rose, 1997), greater co-movement is unlikely to be caused by future increases in trade.

However, both our co-movement variables have the limitation that they have no time series variation (recall that they are defined as the average co-movement over the period 1978 to 1991). As a result, although they can be used to predict which countries will enter EMU, they cannot be used to predict when EMU will occur. Fortunately, since the entry timing decision is highly correlated across countries (i.e. other than Greece the timing of entry is identical across countries), the information lost through not explaining the time series variation in entry decision is minor. In fact, when we use these two co-movement variables as instruments for EMU membership and exchange rate volatility (both defined as above) we find very similar results to the OLS regression (see column marked IV in Table 3) – though the standard errors attached to these estimates are larger.

Once again, we tested the robustness of these estimates through both alternative specifications of instruments (for instance introducing some limited time variation by splitting our averaging period into two sub-samples) and through alternative specifications of the underlying model (e.g. a fixed effects version of the model). None of these changes had a significant impact on the estimated coefficients (see the web appendix for details).

Some may find it surprising that our estimates using instrumental variables are so similar to the OLS version, since many would expect the OLS estimate to be significantly biased upward (arguing that countries with potential for significant growth in trade are more likely to form currency unions). However, as Barro and Tenreyro (2000) point out, the biases could go either way, and their later study (Tenreyro and Barro, 2003), actually finds stronger currency union effects on trade once the currency union dummy has been instrumented. The fact that we do not get the same result probably relates to our more homogeneous sample (no developing countries or high inflation countries) so that the endogeneity effect is only minor in the case of EMU. It is also worth noting that although our two sets of estimates are similar, Durbin-Wu-Hausman tests of endogeneity indicate that instrumentation is appropriate.

The fact that our instrumental variables estimate suggests that it is membership of EMU that is responsible for almost all the increase in trade within EMU (rather than vice versa), need not mean that all of the trade impact is due to the mere act of entering EMU. It is possible that countries that expected to enter EMU altered their policies in ways that encouraged more trade (thus the trade impact is still caused by EMU, but only indirectly). While it is difficult to unravel the direct from the indirect impact of EMU, it is interesting to note that if we estimate our trade model with time dummies just for the EMU-in countries, it appears that EMU is already affecting trade well before EMU actually occurs (as far back as 1994 for the fixed effect version of the model and 1998 for the Rose specification – see appendix). This suggests that the trade impact of EMU is more than just the impact of the single currency – the policy preparations seem to have influenced trade as well.

Although our approach is independent of and very different to that of Micco, Stein and Ordoñez in this volume (i.e. our approach focuses more on the endogeneity issue), it is instructive to compare our results with theirs since the question we address here – namely the trade impact of EMU – is the same. Using a comparable, but shorter, data set, they also find a significant impact of EMU on trade. However, the estimated coefficient in their preferred fixed effect specification is significantly smaller than ours (ranging from 9% to 20% for the most comparable sample), while using a Rose-type specification their results are very similar to ours (an EMU effect ranging from 21% to 37%). Experimentation using a shorter data sample (see appendix) reveals that when we reduce our sample to the period 1993 to 2002, the fixed effects version of our model gives a lower estimate of the EMU effect (9%) while the Rose specification gives very similar estimates (30%). Therefore, it seems that their shorter sample (1992–2003) explains the difference between the results of Micco, Stein and Ordoñez and those presented here. Given that the analysis discussed above suggests that the EMU effect on trade began as early as 1994 it is likely that some of the EMU effect is subsumed into country fixed effects in a shorter sample. However, it is also possible that the lack of a dynamic specification introduces more biases into a longer time series such as the one we use.

3.3. The trade effect of being out

Armed with these results, we can now estimate the trade impact of being out of EMU. Table 4 shows estimates of the possible increase in trade the non-EMU EU countries could have had with EMU countries had they been in the union. The exchange rate volatility effect is estimated by calculating the predicted increase in trade that would have occurred if exchange rate volatility had been zero since 2001 (Greece's entry date).

View this table:
Table 4 Implied trade effect of entering EMU (average increase in trade with EMU member states as a result of joining EMU)

Even though our estimated currency union effect is significantly smaller than that estimated by Rose, the results are still dramatic. For the UK in particular (which has experienced the highest exchange rate volatility of the three) the estimated trade impact is very large.

At first sight these results suggest an enormous trade cost of not being a member of EMU but two caveats need to be kept in mind.

First, this model only estimates the trade impact with EMU countries. If much of that impact is simply trade that has been diverted from non-EMU countries whose comparative advantage is almost identical to EMU countries then the economic impact might be relatively small. However, we have some evidence against trade diversion. If we assume that EMU has no impact on trade between EMU outs, then trade diversion would imply that the creation of EMU would see a decline in trade between EMU-ins and EMU-outs relative to trade between EMU-outs. Estimating our model with an additional dummy variable for trade between EMU-ins and EMU-outs we find no significant impact.

The second caveat is that there is no simple link between trade and overall GDP. This is more difficult to answer and a survey of the literature on this question would merit a paper in itself. However, in a careful review of the evidence HM Treasury (2003) suggests that ‘it seems reasonable to assume that each 1 percentage point increase in the trade to GDP ratio increases real GDP per head by at least 1/3 per cent in the long-run’. Using the figures from Table 4 (including exchange rate volatility effects) and converting the rise in trade with the euro-area to a share of GDP gives a long-run GDP gain of about 7% for the UK and Sweden, and 6% for Denmark.6 Although this is only a one-off effect on the long-run level of GDP, it is not insignificant.

4. FOREIGN DIRECT INVESTMENT7

The impact of monetary union on levels of FDI has become one of the key issues in the economic debate between those who see major advantages in the ‘outs’ adopting the euro and those who see more costs than benefits. In the UK – which has received a disproportionate share of FDI coming from outside Europe – the advocates of entry have highlighted the recent fall in the scale of inflows.

‘With Britain outside the Euro zone it will be increasingly difficult to attract (foreign firms) . . . because they can avoid exchange rate risk on the bulk of their European sales by investing instead in one of the 12 euro countries. This will strongly affect US and Japanese investors who currently invest in the UK as the gateway to Europe, and therefore want Britain to join the euro . . . Britain's share of foreign direct investment coming into Europe has fallen by half . . . We should not be complacent. Britain is not the outstanding place to invest that anti-euro propagandists portray’ (Layard ., 2002).

In this section we first consider the possible links between the scale of FDI and adoption of the single currency. We then consider the potential benefits of FDI. Then we take a careful look at the latest data and draw some conclusions. Here we focus on the ins and the outs among the EU 15. Given the significance of this to the economic debate on entry for the outs – particularly in the UK – it is important to draw out what we can, and more significantly what we cannot, infer from recent trends.

To assess the impact of being an ‘in’ or an ‘out’ on FDI we need to know what motivates such flows. There is a huge literature on the determinants of overseas investment by multinationals. (For good recent reviews see Markusen, 2002 and Lipsey, 2000, 2002). Most FDI flows are from rich, developed countries to other rich developed countries with levels of skills among the workforce which are comparable. Markusen stresses the role of imperfect competition and the desire of firms with information advantages to protect their assets (which include benefits of a good reputation) by setting up production rather than licensing. Such incentives are greater when costs of exporting from the home country are greater. Lipsey stresses the significance of efficiency advantages for some multinationals. He argues that a main motivation of FDI is that it transfers assets and production from less efficient (domestic) owners to more efficient (multinational) ones. ‘Inward investment can be viewed in the recipient countries as freeing capital that had been frozen in industries that the owners would prefer to leave’. Neither of these strands in the literature places a great emphasis upon exchange rate variability or the costs of exporting from a host country to other (third) countries. Indeed as De Menil (1999) points out, the impact of bilateral exchange rate variability on FDI inflows is theoretically ambiguous.

4.1. Currency risk and FDI

The impact that adoption of the euro has on the scale and direction of FDI into different European countries depends crucially on the answers to two questions.8 First, what is the impact of uncertainty about exchange rates between European countries on the location of investment crossing national boundaries? Second, can exchange rate uncertainty be hedged over the time scales that are relevant for long-term investment decisions?

The second issue is relatively clear: for investment projects where cash flows will be generated for many years into the future the ability to hedge any currency risks (for example, those that might occur if costs are incurred in one currency and revenues from sales incurred in another) is severely limited. Foreign exchange forward contracts can be used to hedge risk one or two years ahead – beyond that markets are thin. Whether that reflects some intrinsic problems in long-horizon hedging or an absence of demand for long-term hedging is crucial and really takes us to the more fundamental issue. But for whatever reason it seems highly likely that for those who now do want to hedge currency exposures over periods stretching five, ten or more years into the future the scope to do so is very limited. Furthermore the scale of the risks to be hedged is itself rarely known. Only a company that knew its scale of production, and of costs and revenues and where those revenues were going to come from, would know how much to hedge.

It does not follow from this that the existence of currency risks necessarily reduces FDI nor that its reduction (by a country adopting the euro) increases FDI. In the first place it is not clear how harmful currency volatility is. Risk averse investors in a publicly quoted multinational will care about the correlation between the stock price of the company and their consumption, which in turn will reflect the covariance with other elements of wealth. Even if significant investment by a US or German company into Sweden or the UK makes profits exposed to pound-euro or krona-euro exchange rate volatility it is not clear that this is a risk shareholders should worry about. It may be an entirely diversifiable risk.

One simple way to assess how damaging to shareholders in country X is exposure to unpredictable swings in the real exchange rate of country Y, is to calculate the correlation of the stock returns on a diversified portfolio of country X companies with swings in the country Y real exchange rate. Table 5 shows the correlation between monthly (local currency) returns on national stock indices and the percentage change in the real exchange rate of various other countries. The first element in the matrix is the correlation between the dollar monthly returns on the S&P 500 index and the change in the UK pound real effective exchange rate index. That correlation is negligible (0.001 over the period January 1980 to December 2001). This suggests that investors in US multinationals – who we implicitly assume are primarily US investors holding a major part of their risky wealth in US stocks – should not be much concerned about any exposure they get to UK real exchange rate risk stemming from a decision by a US multinational to locate investment in the UK as a base for possible re-export (hence generating some exposure to pound exchange rate risk).

View this table:
Table 5 Stock market – exchange rate correlations

It is clear that the great majority of the elements in the matrix are trivially small – the only non-trivial elements are the exposures of French and Italian investors to German exchange rate risk. UK exchange rate risk should not be very important to investors who are largely exposed to the US stock market, or to the Japanese market or to continental European markets.

How exposed profits from FDI into a country are to fluctuations in the value of that country's currency will itself depend crucially on the motives for FDI. If the great majority of FDI inflows are for production to satisfy the domestic market of the recipient country then the removal of currency risk will more likely reduce inflows than increase them. After currency risk has gone, overseas companies can locate production for sale into the original recipient of FDI in other, potentially lower cost, areas of the currency union. This might be offset by greater inflows of investment for production for export; but for a country without a cost advantage over others such extra inflows might be limited.

If, however, the larger part of FDI inflows into a country are for production of goods to be exported to other regions then the removal of exchange rate risks with those other regions brings an increase in incentives to inward investment (assuming of course that exchange rate risk itself is costly). But the correlations shown above should make us sceptical about how great those risk costs are. And once again it is relevant that attempts to find a significant exchange rate effect on FDI have not yielded robust results.

This obvious theoretical ambiguity in the direction of the impact of changes in exchange rate risk on the scale of FDI is not easily removed by looking at data on FDI. Breakdowns of FDI by broad sector – manufacturing, services, primary sector – offer few clues as to whether production is primarily for the domestic market or for export. Barrell and Pain (1997) find evidence that a 1% rise in the net stock of inward FDI in the UK, Germany, France and Sweden increases exports by about 0.15% implying that inward FDI does significantly boost exports. But neither theoretical nor empirical evidence on the impact of exchange rate uncertainty on overall investment levels is conclusive. Some recent papers (e.g. Davis and Byrne, 2002) find some negative correlations. But De Menil (1999), in a careful analysis, finds that there is actually a positive impact of bilateral exchange rate volatility on bilateral FDI flows.

4.2. Should we really pay lots of attention to FDI anyway?

Suppose, however, it were the case that FDI inflows were to be significantly and negatively affected by a decision to remain outside the European monetary union. Should that be an important factor behind the decision to adopt the common currency? To some the answer is self evident – surely inward investment is an unambiguously good thing. In fact this is not so obvious. First, FDI inflows are not equivalent to extra investment. FDI inflows represent an increase in foreign ownership and control of assets in a domestic country. That could reflect the change in ownership of an existing asset and not be associated with any new (tangible) investment. (In practice a significant proportion of FDI flows finance mergers and acquisitions.9) Empirical evidence suggests that there is little relation between the scale of inward FDI into a country and the level of capital formation. Lipsey (2000) finds that, if anything, there is a small (and insignificant) negative relation between capital formation ratios and inward FDI ratios. He concludes that the net flow of FDI does not appear to have any significant effect on aggregate capital formation ratios.

The large empirical literature on the impact of FDI has, in fact, not concentrated so much on whether inflows boost the size of the domestic capital stock but rather on whether productivity in foreign-owned firms is higher and whether there are positive externalities from large inflows of FDI. Externalities may come in a variety of ways – greater competition, better training of workers, domestic firms learning from more advanced production methods used by foreign multinationals.

The evidence suggests that labour productivity in foreign-owned firms is indeed higher than for domestic firms. A series of studies shows that foreign-owned firms in the UK have higher productivity than UK-owned firms (see Griffith and Simpson, 2001; Oulton, 2000; Davies and Lyons, 1991). Doms and Jensen (1998) find a similar result for the US: foreign establishments in the US have higher productivity than all US plants lumped together. But both Doms and Jensen, for the US, and Criscuolo and Martin (2002), for the UK, find evidence that the real difference is not so much between foreign-owned and domestic firms but rather between multinationals and domestic producers. The UK evidence is in fact somewhat subtler. Criscuolo and Martin find that non-US multinationals operating in the UK have no higher productivity than UK multinationals operating in the UK. But US multinationals seem to have some intrinsic advantage wherever they set up in the world.

Whether it is nonetheless simply a multinational effect or an FDI effect, the scale of the productivity differences do seem to be large. Griffith and Simpson (2001) find that foreign-owned firms have productivity around 24% higher than the average UK firm; Criscolou and Martin estimate the difference to be 22%.

But the fact that foreign-owned firms seem to have higher productivity than the average domestic firm does not in itself show that FDI is highly beneficial or that losing some part of FDI inflows because of foreign exchange risk is a significant cost. First there is a sample selection issue. It may be that foreign multinationals chose to acquire a stake in domestic companies which (already) have above average productivity. In this case the fact that inward FDI looks to be of unusually high productivity is illusory. This point only has force for that part of FDI which is the acquisition of existing assets and not net new investment. But since mergers and acquisitions typically make up over half of FDI this is still a large part of inflows. Below we present evidence that takes this sample selection issue into account.

Second, the evidence in Doms and Jensen and in Criscuolo and Martin suggests that the productivity advantage of foreign-owned firms is essentially a multinational effect rather than an FDI effect. So ifdomestic multinational firms undertake more domestic investment when foreign multinationals undertake less, overall productivity may not be affected. But given the estimated size of the difference in productivities between the average of domestic firms and the average of foreign affiliates it would require that all of any reduction in FDI was compensated for by investment by domestic multinationals. This is highly implausible. So even if there are no additional benefits from FDI beyond that such investment had higher productivity than investment by average domestic firms there would still probably be some losses from losing FDI. Such losses would be greater if there were in addition positive externalities from FDI.

The literature on positive externalities from multinational companies is a large one. Early work (e.g. Caves, 1974; Globerman, 1979; Blomstrom, 1989; Barrell and Pain, 1997; and Borensztein , 1998) suggested that externalities were large and positive.10 More recent studies (Harrison and Aitken, 1999; and Griffith , 2001) suggest much smaller externalities. Indeed Harrison and Aitken find that negative effects on domestic firms roughly match direct benefits from FDI leaving the overall impact neutral. But their research is based on analysis of FDI into Venezuela and might have limited relevance to FDI into Europe. Baldwin (1999) find that higher penetration levels of inward FDI led to more rapid growth in labour productivity; this is consistent with some positive technology spillovers.

Particularly relevant is a recent careful study by Haskel (2002). They find some evidence of positive impacts upon domestic firms of foreign presence in an industry. Based on the productivity performance of a very large sample of UK manufacturing firms they conclude that a 10% increase in foreign presence within an industry has a positive effect on the total factor productivity of other, domestic firms of about 0.5%. This is a modest, but statistically well-determined, effect. Having analysed what has happened to FDI flows for the ‘ins’ and ‘outs’ we will use the Haskel results to estimate the potential scale of costs to the ‘outs’ of lower FDI.

4.3. Recent evidence on the scale of FDI

Within Europe the economic significance of FDI is generally high but very variable across countries. Table 6 gives an indication. The stock of inward FDI as a proportion of GDP varies widely from over 100% in Belgium and Luxembourg to only around 10% for Italy. The UK has the largest stock of inward FDI (close to 20% of the EU total) though employment by foreign affiliates is relatively low as a proportion of total employment, suggesting that in the UK FDI is unusually heavily concentrated in highly capital-intensive sectors with high labour productivity. The (unweighted) country average of FDI as a percentage of GDP in Europe in 1999 was about 28%; the averages of value added by foreign affiliates as a fraction of GDP and of foreign affiliates’ employment in the whole economy were 16% and 9% respectively.

View this table:
Table 6 FDI Inward stocks in EU 2001

Around one-quarter of all FDI stocks in the European Union are currently outside the euro zone, that is, in the UK, Sweden and in Denmark. Of that fraction the overwhelming part is in the UK. The evolution of the stock of FDI in the three ‘out’ countries is shown in Table 7. Their share has been volatile. It has fallen from around 26% in 1998 to 24% in 2001. But in 1995 it was lower than it was in 2001. But provisional figures suggest that the ‘out’ share fell further in 2002.

View this table:
Table 7 Shares of total stocks of inward FDI in EU in UK, Sweden and Denmark

Table 8 shows the shares of inflows of FDI into all European countries that have come to the three ‘outs’. Once again there is substantial volatility from year to year. But here the downturn post-1999 is clearer. In 2000 and 2001 around 22% of inflows went to the ‘outs’– the share had averaged around 35% in the period from 1985 to 1995.

View this table:
Table 8 Shares of total flows of FDI in the EU in UK, Sweden and Denmark

That part of the flows of inward investment to any EU country that comes from outside the EU is described in Figure 1 and in Table 9. The share of inward investment from outside the EU going into the countries that formed the monetary union in 1999 increased markedly from that date. Between 1999 and 2001 the share of investment from outside going into the euro zone averaged about 70% of all inflows to the EU; in the previous 10 years the average was 54%. The share of inward investment from outside the EU into the UK fell sharply in 1999 to around 25% and has stayed at around that level. But UK investment as a share of all FDI inflows into the EU had been even lower in 1994 and exhibits huge variability. Inflows into the other EU countries that have not yet adopted the euro are also down: in Sweden FDI inflows had averaged 7% of all inflows into the EU from 1989 to 1998 and have fallen to 5% since then. In Denmark the average fell from 3% to 2%. The share of inward investment into the EU going to the three ‘outs’ averaged 31% in 1999–2001 – lower than at any time over the past 15 years with the exception of 1994.

Figure 1

FDI inflows from non-EU countries (€billions)

Source: Eurostat.

View this table:
Table 9 Shares of inflows of FDI from outside EU into EU

There has been less of a decline in the share of intra-EU FDI inflows coming to the ‘outs’. Table 10 shows that the UK received an average of around 13% of all intra-EU FDI inflows between 1999 and 2001, though it has been on a downward trend over that period. The average over the period 1995–8 was just under 12%.

View this table:
Table 10 Shares of intra-EU 15 inflows of FDI: selected countries

Provisional figures from UNCTAD suggest that in the first half of 2002 the UK share of total FDI inflows to all EU countries fell to only just over 5%– down from around 17% in 2002. The share to France was 19% and to Germany 18%– both shares up on their 2001 levels. Eurostat figures show that in the first three quarters of 2002, UK FDI inflows were running at about one-half the level of inflows into Germany and only about 40% of inflows into France. Provisional OECD figures for the whole of 2002 confirm this picture. They suggest there has been a big fall in inflows into both the UK and Denmark. UK and Danish inflows in 2002 are estimated to have been only around half the level of 2001 – and flows in 2001 were themselves only half the scale of flows in the previous year. In France and Germany there was no fall in FDI inflows in 2002. But data from Ernst and Young for the whole of 2002 suggest that the number of projects coming into the UK may only have fallen by 5%. The number of UK inward investment projects remained the largest in the EU at 19% of the total.

4.4. Assessment

It appears that there has been a fall in both the absolute and the relative amount of FDI coming into the UK in recent years, but FDI has historically been highly volatile so it is hard to say how significant is this recent development. The picture in Sweden and in Denmark is even less clear-cut. How should we interpret this? Is the decline in inflows into the UK (relative to the rest of Europe) an EMU effect or just a reflection of the strength of sterling for much of the period since 1999 and perhaps the weakness of US investment, which has historically gone disproportionately to the UK? If the latter factors are the real ones they may prove temporary. Even if they do not they are not single currency factors per se– though of course to the extent that overvaluation of sterling could be removed by adoption of the euro there is a link.

Barrell and Pain (1997) report estimates of the elasticity of stocks of FDI to relative labour costs (between host and investing country). They estimate the elasticity of outward FDI stocks from the UK to be around 0.5 in the short run and 1 in the long term (Table 2, p. 1774). They report estimates of the elasticity of the stock of outward investment from Germany to relative labour costs of around 0.2 in the short run and around 0.33 in the long run. In the UK the average real effective exchange rate between 1999 and end 2001 was approximately 12% above the average over the period 1980–2002 and about 20% above the average for the period 1992–8. Suppose we take the degree of deterioration in relative labour costs in the UK between the mid 1990s and the period since 1999 as 15%. If we use the lower of the Barrell and Pain estimates of short-run relative cost elasticity for the stock of FDI (i.e. the 0.2 figure reported for Germany) we would estimate that the stock of UK FDI should have fallen by about 3%. If we use the higher cost elasticity figure of 0.5 the fall would have been about 7.5%. Between 1998 and 2001 the share of total inward FDI stocks in Europe that are in the UK fell from 21.1% to 18.8% (Table 7). This is a fall in the UK share of about 11%– i.e. 100 × (18.8/21.1 – 1). So a substantial proportion of the fall in the relative amount of FDI in the UK – perhaps around one half using the average elasticity estimates of Barrell and Pain – could plausibly be a result of a, perhaps temporary, rise in UK relative costs rather than as a reaction to failure to adopt the euro.

What about the weakness of FDI flows out of the US in recent years? Might this, rather than EMU, account for the rest of the weakness of inflows into the UK? US inflows into Europe fell by about 40% in 2001 and in the first half of 2002 were down by a further 12% from 2001. So on the face of it weakness in UK overall inflows is perhaps not surprising.

But Table 11 reveals that this explanation does not really add up. Since 1999 there has been a fall in the share of US FDI inflows coming into Europe going to the UK. In 1999 the share was 55%. But in 2001 that share fell to 28% and in the first three quarters of 2002 was still only around 30%. This is against an average for the period since 1991 of around 40%.

View this table:
Table 11 US FDI flows into Europe

So the notion that the very recent fall in the share of FDI coming into the UK is largely due to the weakness of US FDI does not hold up since the share of weak US inflows coming into the UK has itself fallen.

Overall we find that while the competitiveness story for the weakness of UK FDI is likely to account for a significant part of the recent flows it still leaves a lot to be explained. The idea that UK weakness is just the mirror image of low US inflows does not stand up. We conclude that there has been a marked relative weakness of FDI inflows into the ‘outs’ since 1999. Though it is far from clear how much of that weakness is due to not adopting the euro we believe that some significant part may be. But even if that is so does it really constitute a big argument for adopting the euro? The real issue is what is the loss to the economy from receiving less FDI.We now present some estimates of an upper bound on that loss based on UK data.

4.5. The costs of lower FDI

The evidence we described above suggests that the inflow of FDI into the UK over the first three years after the euro came into existence has been only about one-half what might have been predicted given overall inward FDI into all EU countries and based on the UK share over the previous 20 years. One extreme assumption is that all of this ‘lost’ inward FDI was due to not adopting the euro. We now take that as an upper bound on the size of any reduction in inflows and try to estimate its effect on the economy.

First we stress again that there is little evidence that any lost inward FDI would reduce the size of overall domestic capital formation –Lipsey (2000) shows that there is no link between the scale of capital formation and flows of FDI. So any significant impact on domestic production, wages and productivity of changes in FDI inflows will come about as a result of productivity spillover effects, or because foreign assets have higher productivity or both. The second effect (the potential higher productivity of assets owned and controlled by foreign multinationals) may be overestimated by simply applying the average excess of productivity and wages in foreign firms over domestic firms, since foreign firms may choose to acquire domestic assets that were already of higher average productivity. So the 25% or more average productivity advantage of foreign firms is not a good estimate of the true enhancement to productivity. This is a sample selection/bias issue analogous to that which arises in the trade issue we considered above.

Fortunately there is some very good evidence for the UK that allows us to adjust for the potential impact of this bias. Conyon (2002) analyse panels of firms in the UK taken over by domestic and foreign acquirers. They also used matched data for over 600 firms that were not taken over and use difference-in-difference techniques to estimate the productivity effects from assets simply coming under the control of a foreign multinational. They find that the differential effects on wages and on productivity are substantial. Having controlled for a wide range of factors, they estimate that the pure productivity enhancing effect of assets being managed by a foreign acquirer is to increase productivity by 13%. Around half of the higher productivity is reflected in higher wages paid by firms that come to be owned by foreign multinationals.

We take a 10% figure as a central estimate of the direct productivity effect of foreign presence.11 We add to this the estimate by Haskel et al. of the positive spillover effect of foreign presence on domestic firms. Their estimate is of an elasticity of TFP to foreign presence of approximately 0.05 (a 10% rise in foreign presence raises productivity of domestic firms by 0.5%). We take an upper estimate of the induced fall in FDI flows into the UK as a result of not adopting the euro to be that flows are reduced by 50%.

An extreme assumption is that outside the monetary union the flows into the UK will remain at the same relative level as seen since 1999. Combining these estimates and assumptions we now estimate an upper bound for the impact upon wages and productivity of not adopting the euro. The share of value added from foreign affiliates in the UK in 1999 was just over 4%. Had the UK share of inward investment always been at its 1999–2002 value the stock of foreign capital might have been half that size and foreign affiliates contributed only 2% to GDP. If foreign-owned firms have a 10% productivity advantage the direct effect of lower foreign presence on productivity is then 10%× 2%= 0.2% (i.e. firms that produce 2% of GDP would not be owned by foreign multinationals and would as a result have a level of productivity 10% lower). To this we add the lost benefits of positive spillovers based on an assumed elasticity of 0.05. This generates a further loss of 2%× 0.05%= 0.1%.

The overall loss is 0.3% of the level of productivity. This is a long-run, steady-state loss and is based on the assumption of permanently losing half of FDI flows as a result of not adopting the euro. The scale of losses would be far lower than this in the initial periods after not adopting the single currency and would only gradually build up to that level. At current values the long-run level of losses, only experienced after many years outside the single currency area, would be about $5 billion a year in the UK.

That is not a trivial figure. But neither is it large and we consider it very much an upper bound on the long-run cost. It is in any case much smaller than the potential trade effects we discussed in Section 3 above. On the basis of this simple calculation, to base a decision on adopting the euro largely on the possible impact of FDI would be to give it far too much weight.

5. FINANCIAL MARKETS: ACTIVITY AND LOCATION

The introduction of the single currency raised the possibility that the centre of financial activity in Europe would shift from London to Frankfurt. Currency unions, and time zones, typically have a single dominant centre, usually located in the same city as the dominant currency's central bank. With the ECB located in Frankfurt, and London being outside the union, the latter's status as Europe's financial centre was vulnerable. A crucial issue for London is whether it can remain the principal location for euro-denominated financial activity despite not being part of the euro area, particularly if the euro's importance as an international currency grows. The equivalent issue for the ins is whether they will continue to see the bulk of activity in their own currency conducted by one of the outs.

London has many advantages as a financial sector to set against those of Frankfurt including a critical mass of activity; high-volume clearing systems; a high concentration of financial firms, support services, physical assets, labour, and legal and accounting expertise. It also has the advantage of having English as its common language. The probability that the UK will eventually adopt the euro must diminish one of Frankfurt's advantages, leaving it with just the proximity of the central bank in its favour. Even this is diluted to some extent by the presence of the Bank of England, which would probably have a higher profile in the ESCB were the UK to join. It is significant that most of London's business is denominated in foreign currency (mainly in euros and dollars), while that of Frankfurt and Paris is denominated in local currency.

5.1. Evidence

There is little evidence to suggest that use of the euro has increased relative to that of its legacy currencies since 1999. For example, while the euro is the second most used currency in foreign exchange transactions, its use is only slightly greater than that of the DM in 1999.12 This pattern is also apparent in international banking, over-the-counter interest rate derivatives, insurance and official reserves. The euro has increased its profile in two areas, however: there has been a significant increase in euro-denominated credit derivatives (although demand for these appears to have come from Europe rather than the US or Japan), and in corporate bonds where the euro's share increased from 29% in 1998 to 36% in 2001. Not surprisingly, activity in euro-denominated government bonds has declined with the introduction of restrictions on European government borrowing. Significantly, however, London's share of euro-denominated Eurobond issuance has increased since 1999, from 50% to 60%.

In terms of size, London is clearly the dominant player in Europe (see Table 12), and there is little sign so far of its share of international business declining, as Table 13 shows. The only area of financial activity in which Frankfurt has grown significantly is in exchange-traded derivatives. Although London appears to have suffered no ill effects from being outside the euro area, there are signs, however, that the smaller centres in Europe are losing euro-denominated business to Frankfurt, Paris and, significantly, London.

View this table:
Table 12 Shares of international financial business, 2002 (in %)
View this table:
Table 13 Recent changes in London's share of international financial business (in %)

Overall, while four years would probably be sufficient time to see some evidence of a migration from London had EMU significantly altered market participant's location decisions, we can find none.

6. MACROECONOMIC PERFORMANCE

While some of the economic consequences of monetary union, such as changes in FDI, may become apparent quite quickly, others, such as improvements in growth, unemployment and inflation, may take many years to emerge. Although questions about macro performance are not unreasonable even at this stage of the union, as recent experience in Germany has shown, we do not yet have enough data to warrant a detailed econometric investigation. In this section, therefore, we report the results of a very simple, broad-brush search for indications that macroeconomic activity has responded to the single currency.

6.1. Has the performance of the outs differed from that of the single-currency area as a whole?

When the euro was introduced most of Europe had been on much the same macro trajectory since the trough of the early-1990s recession in 1994. The recovery after that was prolonged, remarkably smooth and characterized by declining inflation accompanied by increasing output. Significant changes to the structure of macroeconomic policy also took place in this period as newly independent central banks took aim at inflation targets, and abandoned, or at least downgraded, the use of intermediate targets such as monetary growth. The broad approach to fiscal policy also changed as governments attached increasing weight to balancing their books.

The trends of the previous six years were interrupted in mid-2000 by a negative demand shock worldwide, accompanied by monetary tightening, and fiscal loosening, in most European countries. European economic performance reacted to these changes in a fairly uniform fashion. Growth increased in the first year of the euro, declined dramatically in 2000 through 2001, and began a slow recovery in 2002. Unemployment and inflation too have displayed fairly uniform behaviour across Europe: unemployment has generally declined continuously since 1998, while inflation increased in 1999 towards a peak around 2000, before declining to what appears to be a fairly stable level for 2002, at or above the levels of 1998.

On the basis of data from 1998 to 2001, only a small number of exceptions to the general trends stand out. In particular, the major exception to the common-growth story has been Norway, where growth and unemployment have deteriorated since 1998. The other major trend outlier is Swedish unemployment, where the fall of 3.4% since 1998 considerably exceeds the 2.2% of the euro area. In terms of shocks, the UK appears to have responded best to the slowdown in 2000 (with the arguable exception of Norway), with its growth rate falling by just 0.9% while the euro average fell by 1.4%. In terms of inflation the UK is also slightly out of step with the euro zone in having a rate that was lower in 2002 than in 1998. However, even these differences are minor and cannot easily be attributed to being in or out of EMU, except to the extent that the UK was able to run a tight monetary policy before 2000 and an expansionary one thereafter – something it could not have done as an in.

6.2. Have the outs performed better than individual members of the single currency?

While there appears to be no clear evidence that the outs have performed better or worse than the ins as a group, it is possible that some significant differences might be present on a bilateral basis, since economic performance since 2000 has not been uniform across all the members of the single currency. In Table 14 we present the results of bilateral comparisons of performance in terms of growth, unemployment and inflation. For each country we calculate the average values of each variable for the periods 1995–8 and 1999–2001. We then compare these average levels across countries in each period. Finally, we present a comparison of the change in the average levels across the two periods for each country. In each case we award a score of 1 to the ‘winning’ country, and enter the sum of all these scores along the rows of the table. So, for example, the first panel of Table 14 shows that Denmark had a higher average level of growth than five of the ins in 1995–8.

View this table:
Table 14 A scorecard of macroeconomic performance

In terms of growth the ins clearly score better than the outs taking the two periods together. From a total of 83 bilateral comparisons, the outs come out better in only 27. In terms of the improvement in growth, however, the scores are about even; the outs scoring 20 out of 39. The picture is reversed for unemployment. The outs perform better in 45 of 55 comparisons in 1995–8, and 38 of 50 in 1999–2002. The outs also do better in terms of inflation performance, scoring 35, 42 and 28 out of 55 for the three comparisons.

Overall, taking only the comparisons that involve the EMU period, the outs score 35 out of 80 for growth, 64 out of 100 for unemployment, and 70 out of 110 for inflation. Thus in 169 cases out of 290, i.e. 58%, of the bilateral comparisons the performance of the outs has either been better than that of the ins since 1999, or has improved more in comparison with the pre-euro period, or both. Thus there does not appear to be any strong evidence that the ins have performed better than the outs so far.

7. CONCLUSIONS

Currency unions often last many decades – good ones last many centuries. Can we learn anything from the first four years of the European currency union? In terms of the overall macroeconomic performance of the EU ‘ins’ relative to the performance of the EU ‘outs’ there is no killer fact. The outs have, on balance, done slightly better on some things than most of the ins – the ins have generally done slightly better than the outs on a range of other performance measures. But the similarities in overall performance are more striking than the differences. Inflation in most European countries has been low and unemployment has fallen since 1999, net, in most countries.

Not surprisingly, in broad macroeconomic terms it is not yet obvious what difference the monetary union has made. But in one area where we might expect monetary union to have had a significant and relatively rapid effect – trade between currency union members – there are signs that the Euro has made a difference. We find significant evidence for a trade enhancing effect on bilateral trade within Europe from the formation of the single currency; outs might do a lot more trade with ins if they adopted the euro. On FDI the evidence is less clear – there has been a fall in FDI to the non-member countries, but these flows have historically been volatile for many reasons. In addition, such a fall, even if statistically significant, is certainly much less economically significant than the potential trade effects just discussed. In contrast, we find no evidence that EMU has yet influenced financial market location, and that London's position as the principal financial centre in Europe appears to have been unaffected by its being outside of the euro zone.

The potential welfare gains from the outs adopting the euro as a result of enhanced trade and greater FDI are significant. Against those gains should be set the costs of losing some degree of independence over monetary policy. The scale of those costs depends on the likelihood of there emerging substantial divergence from cyclical positions in the rest of the euro area. Little can be learned about that from the experience of the relatively short period since the launch of the euro.

Discussion

David Begg

The Business School, Imperial College London and CEPR

I greatly enjoyed this paper, which provides an early assessment of what we have learned about the consequences of remaining outside the euro. In some respects, of course, it is too soon to say; but in other respects the evidence is already convincing. Broadly, I share the authors’ judgments as to which questions have been resolved since the start of the euro and which questions will be answered only after we have much more evidence than is yet available.

On the issue of the trade effects of the euro, we have already learned a lot. Before the advent of the euro, pioneering studies of the effects of currency unions had to rely largely on small and tiny countries about whose relevance to Europe one could never be sure. Now we have evidence from European countries themselves. From this we have learned two things. First, the qualitative conclusion that currency unions promote trade has survived the European transplant. Second, in quantitative terms, the effect is probably between 20 and 50%. The lower end of the range is suggested by the companion paper by Micco, Stein and Ordoñez in this volume. The upper end is the ceiling suggested in the recent assessment by the UK Treasury. Barr, Breedon and Miles come down in the middle of this range.

To the extent that such estimates rely on comparing what happened before and after the launch of the euro, this may still prove to be an underestimate of the eventual effect, for two reasons: first, the ‘pre-euro’ data display some effects that occurred only because of the anticipation that the euro was on the way; second, the ‘post-euro’ data are unlikely yet to include all the long-run effects. Hence, the difference between before and after understates the true effect. Barr, Breedon and Miles adopt a central estimate of 29% for the trade-creating effect of the euro, but this may turn out to be an underestimate. Even so, it is much more plausible than the estimates of 200% or more that were derived from earlier data based on currency unions comprising small countries and dependencies of colonial powers.

I note in passing that the UK Treasury's assessment of the ‘5 tests’ for UK membership, published in June 2003, acknowledged that extra UK trade from the euro could be anything up to an annual £1700 in per capita income, or nearly £7000 a year for a family. It is hard to think of any government policy actually being implemented today acknowledged to have benefits this large.

A related issue is whether the formation of a monetary union causes trade diversion that reduces the absolute level of trade between ins and outs. To date, the evidence – whether for Europe or for small countries and dependencies – suggests that no such trade diversion can be detected. For the European outs, this has two implications. First, the formation of the euro zone is unlikely to interrupt the ever-increasing degree of integration that is occurring within Europe. Hence, European outs are likely to adopt the euro sooner or later. Second, by failing to join at the outset, the outs are forgoing trade creation that they could have had.

The welfare cost of forgone trade is much smaller than the level of trade forgone. It must be identified as little triangles of consumer surplus not exploited, as failure to enjoy economies or scale and scope, as profit shifting, as failure to unlock benefits of diversity and competition, or as gains to the greater pursuit of comparative advantage. Any estimate is still subject to a large confidence interval.

The second part of the paper discusses the fall in inward investment to the outs since the start of the euro zone. Broadly speaking, this argument is in three parts: first, investment is down but is volatile, so the confidence interval on any implied reduction is large; second, even if a fall has taken place, our models of FDI are less well developed than our models of trade, so it is more difficult to be sure that the fall should be ascribed to the advent of the euro; and, third, even if it was caused by the euro, the welfare effects may be small.

The UK share of EU inward investment averaged 39% in the decade prior to the launch of the euro, and has averaged 23% in the subsequent three years. Yet it will take several more years of data before we can be sure that there has been any euro effect. The authors argue that some of the reduction in UK FDI should be attributed to the sterling appreciation during 1996–2002. While I agree that this probably inhibited inward investment, I am less sure that it is appropriate to view it as unconnected with the launch of the euro. In particular, if the UK had been the first wave of euro entrants, sterling would have been locked into the euro at a more competitive exchange rate than that which subsequently transpired.

I share the authors’ judgment – a point not always recognized – that, although productivity is higher in firms that are foreign owned, it is difficult to infer from this to what extent higher productivity is caused by foreign ownership and to what extent foreign owners simply acquire firms with above average productivity. Although this is an important caveat, I still suspect that technology transfer and investment in knowledge-intensive industries are of increasing importance, especially for the intra-industry trade that so dominates trade within Europe. Hence, in relation to FDI I conclude that, while the case cannot yet be proven beyond doubt, there is little good news for the outs.

The final part of the paper deals with the comparative macroeconomics of ins and outs. This is a bit like viewing an early lap of an Olympic 10 000m race and trying to make inferences about the likely eventual winner. For many purposes, it is simply too short a span from which to draw any reliable judgment. Yes, Germany has been in recession and has adopted the euro, but do Real Madrid victories in the Champions League demonstrate that the euro is good for football too? The people who argue on Mondays, Wednesdays and Fridays that Germany could have reflated but for the strait of the euro are the same people who on Tuesdays and Thursdays point out that most German problems are microeconomic and supply side.

Thus, while I happily endorse the view that there is scope for improving the structural and institutional design of macroeconomic policy in the euro zone, I think that simple scorecards are not informative.

Jonathan Haskel

Queen Mary, University of London

The paper sets out to answer two key questions. First, how are those countries on the outside of the euro doing? Second, and rather harder, how would those countries currently on the outside do if they became insiders?

As the paper rightly points out, getting sensible answers to the first question is hard owing to the short time period. However, the indications are that, if anything, countries outside are doing rather well, the key example being the UK whose output growth, for example, looks very impressive. The authors point out this was likely a consequence of a monetary policy that would not have been feasible with the euro: tight before 2000 and then expansionary thereafter.

Answering the counterfactual of what would have happened to macro policy if the outsiders had been insiders is comparatively easy since there are more or less specified macro rules. It is rather harder to say what would have happened to trade and FDI. Since the pro-euro lobby regards euro membership as being beneficial for both these flows this is a key question that the authors address.

Starting with trade the essential claim is that membership of a currency union benefits trade, as in the upward sloping line in Figure 2. Thus if we knew the slope of the line we would be able to work out how much extra trade would arise were a new member to join the euro. The problem of identifying this relation from actual data is set out in Figure 3. Actual data presents us with a group of countries who are ‘in’ and a group who are ‘out’. But joining is not a random activity. If countries who are ‘in’ would also have a higher trade relationship anyway (they, for example have good political relations which both raises trade and the chances of being in) then they lie on a higher locus. Hence estimation from observed data will likely overstate the gains from joining, unless this can be corrected for. In a regression framework the problem is that we need to run a regression of

Figure 2

Relation between trade and membership of a currency union

Figure 3

Identifying the relation between trade and membership of a currency union from actual data

Yi=β1EUROi+β2Xi+vi3

where Y is some variable (inflation, FDI, trade etc.), EURO is a dummy for membership and X every other control that might conceivably have any effect on Y and is correlated with EURO. If we omit an element of X the coefficient a1 is of course a biased predictor of what would happen if a randomly selected country joined the euro.

There are of course two solutions to the problems in (3); to control for X or to find a good instrument for EURO. Both are tried in this paper. The first is to estimate a gravity equation with (log) bilateral trade (Y) as a function of euro membership plus exchange rate volatility, output, output per head, distance, contiguity, language and EU membership. This gives a coefficient on EURO of 0.25, suggesting that trade increases by 25% with EURO membership. A fixed effects estimate gives a coefficient of 0.24. The second method, IV, gives a coefficient of 0.21, where the instrument is output and price co-movement. In sum, all of these measures suggest an increase in trade of about 20–25%.

A number of issues arise from this discussion. First, as the authors point out, these are large effects. While it is true that they are less than some of the original Rose effects (Rose, 2000, for example) this may well be because the trade/union relation is concave and much of the variation in the Rose estimates arose from small countries with little trade initially. Note, however, that the overall effect of euro membership would be 72% taking into account the implied reduction in exchange rate flexibility.

Second, I was somewhat surprised that the fixed effects estimates are so close to the non-fixed effects estimates. This is not the case in the Micco, Stein and Ordoñez paper in this volume, where the use of fixed effects involves large falls in β1, from 20% to 6%. Indeed their fixed effects estimates are around 6–9%, rather than the 20–25% here. This fall in the estimates seems plausible to me. For β1 to be downward biased, factors omitted from X which are captured by the fixed effects would have to be those that make trade more likely but joining the euro less likely. It is hard to think of such factors; rather the relation is likely positive making β1 upward biased. Thus 20% as an upper bound seems sensible to me.

The authors discuss the relation between their findings and those of Micco et al. in some detail and document a very interesting finding, namely that restricting their data set from 1978–2002 to 1993–2002 the panel coefficient falls to the much lower 9% effect found in Micco et al., suggesting that it is the short period that accounts for the differences. Given how critical this coefficient has become in the UK public debate on joining EMU (HM Treasury, 2003) this is important and the paper also suggests that quite a lot of the trade effect comes from the anticipation of joining EMU. One problem here is to try to also unravel the effects of the Single Market Programme that was supposed to have been completed in 1992 but might also have had some lagged effects.

Third, and similarly, I was also surprised that the IV and OLS estimates were close together. Recall that IV estimates identify the marginal effect for the particular part of the sample that makes up the instrument. In this case it is the countries with the greater output co-movement whose marginal effect on trade is being identified by the IV result. So can we draw any inference for countries with smaller output co-movement, who from Table 1 are countries currently outside the euro? If the trade/union relation is concave the marginal effect might be greater for the current outs, which would put 21% at an underestimate. On the other hand, if the amount of trade depends on the instrument as well as the included X factors then the estimate is likely biased upwards, since the part of the variation in joining the euro or not that it tries to capture is correlated with trade net of the X factors. This latter possibility seems more likely to me. All in all, the 20% figure looks to be an upper bound on bilateral trade.

Does the evidence on FDI provide more support? The authors start by trying to untangle whether FDI has fallen in the UK while it has been out. This is a hard calculation to make since there is a relatively short time period and there has been a trend shift towards FDI in Eastern Europe since the fall of the USSR. Thus the authors are rightly cautious in concluding there is some indication that the UK is losing out. So is this a good reason to join the euro? There are a number of issues here.

First, as the authors point out, the extent to which currency risk is a current disincentive for FDI to come to the UK depends on the amount of goods that are exported to the EU from UK-located affiliate plants. Direct evidence on this issue for US multinationals is set out in Slaughter (2003) who documents that in 1999 75% of Ireland-based affiliate sales were exported, whereas 75 of UK-based affiliate sales were into the host market. Since the US is by far the largest source of FDI to the UK, other affiliates would have to have hugely opposite figures for the main destination of affiliate sales to be outside the UK. This might explain why the loss, if any, to the UK, of FDI from being outside the euro, is hard to detect.

Second, on the general issue of the desirability of FDI it is important to distinguish between whether foreign firms are desirable simply because they raise productivity via a batting average effect, or whether they might be the source of productivity spillovers. On the batting average effect, if foreign firms are simply more productive, the implied possible decrease in productivity from the loss of FDI follows, depending on the replacement effects there might be from UK firms. If there are spillovers then there may be detrimental effects on productivity directly if FDI is lost.

Regarding the batting average effect, it used to be thought that the observation that foreign firms are more productive was clinching evidence that the loss of foreign firms would reduce host country productivity. With the papers by Criscuolo and Martin (2003) for the UK and Doms and Jensen (1998) for the US, this view has been discredited. A simple comparison of foreign firms operating in country X with domestic firms operating in country X compares foreign firms, who are by definition multinational enterprises (MNEs), with all domestic firms. Since all domestic firms include domestic MNEs and non-MNEs the comparison is not like-with-like. The appropriate comparison is foreign MNEs and domestic MNEs. When one does this, as Criscuolo and Martin (2003) and Doms and Jensen (1998) are able to do, one finds that MNEs generally have about the same total factor productivity (TFP; although US MNEs seem to have higher TFP than everyone), so that the apparent domestic disadvantage arises due to the mixing of domestic non-MNEs and domestic MNEs in the crude calculation. Indeed, Criscuolo and Martin (2003) are able to go further. They document that the source of the MNE advantage, which is in part due to managing plants better, but mostly due to owning better plants in the first place. In sum, the loss of productivity due to the possible loss of FDI and the consequent batting average effect seems likely to be very small.

What about the loss of productivity due to spillovers? Estimates of spillovers from foreign firms are notoriously hard to pin down. Using plant-level data Haddad and Harrison (1993) and Aitken and Harrison (1999) find negative spillovers from increased industry-level FDI for Moroccan manufacturing and Venezuelan manufacturing, which they speculate might be due to unmeasured poor absorption capacity of plants in developing countries. Haskel (2002) and Keller and Yeaple (2003) find positive spillovers using plant and firm level data for the UK and US respectively. The former paper finds that the spillover effects seem rather modest; the 10 percentage-point increase in foreign presence in the UK in 1973–92 accounted for about 5% of the observed rise in UK manufacturing TFP. Let us suppose that the UK lost half of its foreign presence, i.e. a decade's worth of FDI, as a consequence of being outside the euro. It would then have had 2.5% less of its rise in TFP; this seems like a small number.

A counter-argument to this is the rather higher numbers estimated in Keller and Yeaple (2003) who find that positive spillovers account for around 15% of US TFP growth in 1986–97. Their only significant spillover findings, however, are in the US high-tech sector for which they document a strong positive relation between US firm TFP growth and foreign presence. Given that the US is likely the productivity leader in this sector it does seem more likely that this coefficient is exaggerated due to reverse causation, i.e. that foreign firms enter the US high-tech sector to learn from the US rather than the other way around.

In sum, I was not convinced that the losses from being on the outside in terms of lost trade and FDI were serious. Thus I agree with the authors’ conclusion that on these numbers at least, the case for the UK ‘out’ becoming an ‘in’ is not compelling.

Panel discussion

Carlo Favero suggested an explanation for why the coefficient of the trade effect of EMU does not fall much if instruments are used: it is well known in the literature that IV estimation does not affect coefficient estimates if the stochastic trend is dominant. Indeed, compared with the paper of Micco, Stein and Ordoñez the sample contains more time-series observations. He wondered why the authors did not estimate a dynamic model with a lagged dependent variable. Alejandro Micco pointed out that including fixed effects in the regressions does not affect the coefficient estimates because the authors use a longer time period so that endogeneity matters relatively less. Ernesto Stein found it puzzling that the endogeneity bias of OLS estimates might be considered to go in either direction: it is hard to imagine that a country that trades more is less likely to join the currency union. David Miles replied that he agreed that IV estimation should lower the estimate of the currency union effect. Michael Ehrmann suggested that a way of making better use of the data would be to distinguish between those countries that had been more homogeneous pre-EMU and those that had been more heterogeneous.

Jean-Marie Viaene mentioned that the FDI share of euro area countries might be inflated because of a substantial volume of merger and acquisitions (M&A), as for example in Belgium. Thus, the FDI share of the UK and other countries outside the euro area might seem artificially small. He suggested distinguishing between M&A and truly new capital. Hans-Werner Sinn added that the most important components of FDI are typically M&A and retained earnings and not new capital. David Miles replied that the back-of-the-envelope calculations addressed this comment by assuming that the capital stock remains unchanged whereas the ownership changes. Thomas Moutos argued that even if productivity increased with FDI, this would not imply that welfare increases if competition decreases at the same time. Steve Cecchetti asked whether the decrease of FDI in the UK could be explained by legal and regulatory changes in the continental European countries that have decreased the competitive advantage of the UK over time. David Begg considered this as unlikely because such changes had been well underway for quite some time.

George de Ménil thought that FDI in central Europe is not important quantitatively. Alejandro Micco added that the risk of investors with respect to FDI had increased for EU countries outside EMU so that expected profits for such investments had fallen. Karen Helene Midelfart Knarvik asked for a more detailed analysis of FDI as a consequence of EMU. She expected the UK to be affected more for FDI locating in the UK to serve as an export platform as compared with FDI meant to serve the UK market. David Miles replied that such a disaggregated analysis was not feasible because of data limitations. More importantly, it is quite plausible that EMU affects FDI that serves to supply the UK market.

Finally, Ernesto Stein wondered why the authors dismissed a high inflation history as an important criterion for an optimum currency area after EMU. High inflation history still matters as a motivation for countries like Italy to remain in EMU.

APPENDIX: EXPLORING THE TRADE IMPACT OF EMU

This appendix presents a number of robustness checks and alternative specifications of our estimates of the trade impact of EMU.

Country exclusion tests

Table A1 shows the impact on the EMU coefficient of dropping individual countries (and all the EFTA countries) from both the Rose and fixed effects specifications of the OLS trade model. Note that the fixed effect specification drops the variables distance, language and contiguity since they have no time series variation.

View this table:
Table A1 Estimated EMU coefficient after removing countries

In all cases the EMU coefficient remains significant at the 1% level. The largest change comes from dropping Denmark from the Rose specification; this is due to the strong interaction between the EMU effect and the exchange rate volatility effect. It seems that Denmark, as a low volatility non-EMU country, is important in discerning the volatility effect from the EMU effect. Certainly, the fall in the EMU coefficient is mirrored by a rise in the volatility coefficient.

Variable exclusion tests

Table A2 shows the impact on the EMU coefficient of dropping individual variables from the Rose and fixed effect version of the trade model. Given the strong interaction between the volatility effect and the EMU effect in the Rose specification, we also show the impact of dropping further variables after the exchange rate volatility effect has been dropped.

View this table:
Table A2 Estimated EMU coefficient after removing variables

The EMU effect remains significant at the 1% level in all cases except when the distance variable is dropped from the Rose specification. Once again, the interaction of the exchange rate volatility and EMU effects is responsible for this (the volatility effect increases significantly when the distance variable is dropped).

Changes in estimation period

Table A3 shows the impact on the EMU coefficient of changing the sample period used in estimation (the full sample runs from 1978:1 to 2002:1). We show the impact of dropping observations from the beginning of the sample as well as from the end.

View this table:
Table A3 Estimated EMU effect in different sample periods

The EMU effect remains remarkably stable across sample periods for the Rose specification, but declines significantly in the fixed effect specification when earlier observations are dropped. As discussed above, this explains the differences between our results and those of Micco, Stein and Ordoñez in this volume, who use a shorter data sample. All coefficients remain significant at the 1% level.

Specification of instrumental variables

As discussed above, while our instruments are appropriate in the sense that they are correlated with the variable they are instrumenting (EMU entry) but independent from factors that might influence trade post-EMU, they do have the limitation that, in the form used above, they have no time-series variation. Table A4 presents alternative specifications of our two instruments (price co-movement and output co-movement) that introduce some time-series variation. Not only may such specifications help identify whether time variation has an impact on the EMU estimates for the Rose specification, time-series variation is a prerequisite for using our instruments in the fixed effect version of our trade model.

View this table:
Table A4 Impact of alternative instrument specification

We present two alternative specifications of our instruments: First, instead of simply letting the instruments take a single value of the whole time-series sample, we set the instruments equal to zero up until EMU entry (1998:4), and then equal to their 1978–91 averages post entry (1999:1 onwards). The second specification sets the instruments equal to the 1978:1 to 1987:4 average over that period and then to their 1988:1 to 1997:4 average for the period 1988:1 to 2002:1 (i.e. we split the pre-EMU sample in half).

As Table A4 shows, the alternative specifications make little difference to the estimated coefficient. All coefficients are significant at the 1% level except exchange rate volatility in both versions of the fixed effect model in which it is significant at the 5% level.

EMU effect over time

As noted in the text, not all of the impact of EMU need be due to the introduction of a single currency, some may be due to policy changes made by countries that knew they were going to enter, or other anticipation effects. To give some idea of this pre-entry effect we look at the evolution of the EMU effect through time by introducing a whole set of dummy variables that take the value one over a given year for EMU countries. For example EMU93 will equal one for all four quarters of 1993 if the country pair involves two countries that will enter EMU. Figure A1 shows the evolution of the coefficients on these dummies for our two models. In the case of the Rose specification we show the actual coefficients on these dummies, while in the fixed effect version we show the deviation of the dummies from the average coefficient pre-1992.

Figure A1

Evolution of the EMU coefficient through time

In both cases it seems that the EMU effect increases most strongly in 1998, the year before EMU entry. However, for the fixed effect model we begin to see significant EMU effects as early as 1994 indicating perhaps that the preparations for EMU had an effect on trade.

WEB APPENDIX

Available at http://www.economic-policy.org

Footnotes

  • The Managing Editor in charge of this paper was Paul Seabright.

  • 1 See, for example, Layard (2002).

  • 2 The countries are Austria, Belgium, Switzerland, Germany, Denmark, Spain, Finland, France, UK, Greece, Ireland, Iceland, Italy, the Netherlands, Norway, Portugal and Sweden.

  • 3 We experimented with a range of different specifications that gave similar results. For example, a fixed effect version of the model above gave a significant currency union coefficient of 0.24.

  • 4 29% is exp(0.254).

  • 5 Tenreyro and Barro (2003) propose another highly ingenious instrument – namely the joint probability that two client countries adopt the same anchor currency – on the assumption that bilateral trade between the two clients is not influenced by the anchor economy, only by the common currency. Unfortunately, this instrument is not appropriate for our study as EMU has no defined anchor currency, or any alternative anchors.

  • 6 Note that although our results suggest that the UK will get the largest trade boost from EMU, UK trade with the euro area is a much smaller share of GDP than Sweden, which in turn has a smaller share than Denmark, so the overall GDP impact is similar for all three.

  • 7 The authors would like to thank Rachel Griffith and Alexander Klemm of the Institute for Fiscal Studies for helpful conversations on the issues raised in this section and for allowing us to use some data that they had assembled.

  • 8 A third factor is what the impact of joining a currency union is on competitiveness, particularly relative labour costs. This will depend upon the conversion rate at entry.

  • 9 Devereux and Griffith (2002) note that the OECD has estimated that mergers and acquisitions account for more than 60% of all FDI. See OECD (2000).

  • 10 Barrell and Pain estimate that a 1% rise in the stock of foreign-owned capital raises technical progress in German manufacturing by 0.26% and in UK manufacturing by 0.27%. But they find no evidence of a positive impact on stocks of FDI in the UK service sector, and this is where most FDI has gone.

  • 11 We scale down the 13% figure of Conyon et al. to allow for the fact that not all the gains accrue to the host economy – they estimate that wages only rose by around half the 13% figure. We consider 10% an upper bound on the effect on domestic residents.

  • 12 In April 2001 the euro figured in 43% of foreign exchange transactions by value, compared with 85% for the dollar, and 26% for the yen.

REFERENCES

View Abstract