A reconsideration of the optimum currency area approach: the role of external shocks and labour mobility.

Gros, Daniel

1. Introduction

Discussions of the economic costs and benefits of EMU usually take as their basis the optimum currency area approach. This approach starts from the premise that when an external shock hits the economy it is easier to adjust the exchange rate than domestic prices or wages. In the words of Mundell (1961), page 657:

'A system of flexible exchange rates is usually presented, by its proponents, as a device whereby depreciation can take the place of unemployment when the external balance is in deficit, and appreciation can replace inflation when it is in surplus.'

Most economists accept the general idea behind this approach, namely that nominal wages are usually sticky in the short run and that it is therefore easier to adjust to external shocks and obtain changes in the real exchange rate or the terms of trade through a movement in the exchange rate. But there is little agreement on how important these 'external' shocks are in reality. Will the loss of the exchange-rate instrument lead to massive unemployment because large negative external shocks are likely? Or do external shocks play only a marginal role in the evolution of unemployment? The presumption of most economists would be that external shocks should have a significant impact, at least for small countries.

The first key issue for an evaluation of the economic case against EMU is thus: do external shocks (i.e. shocks to exports and/or the exchange rate) have a strong impact on (un)employment in member countries? This is an empirical issue that has not been addressed in the literature up to now.

A second key point of the optimal currency area approach concerns the role of labour mobility. In the words of Mundell (1961), page 661:

'The argument for flexible exchange rates based on national currencies is only as valid as the Ricardian assumption about factor mobility.'

The latter has two aspects according to Mundell:

'that factors of production are mobile internally, but immobile internationally.'(ibidem).

The emphasis on the difference between inter-regional and inter-national labour mobility in Mundell is often overlooked in discussions about EMU. If one were to find that labour mobility is as low within member countries as it is between them one would have to conclude (yet again!) with Mundell (ibidem p. 660) that:

'The optimum currency area is the region'

To paraphrase: the case for flexible exchange rates based on national currencies is only as strong as the difference between inter-regional and inter-national labour mobility.(1) This key point is almost always overlooked in the literature on EMU which considers only one aspect namely the low degree of international labour mobility within Europe. It is assumed that mobility within a country is higher than that between countries.

But even if one abstracts from this argument the larger question remains: How important is labour mobility in theory and in practice? Is the general impression that labour mobility is extremely low in Europe justified? Is more labour mobility desirable for EMU (because it facilitates adjustment)? Or is it undesirable (because it favours concentration), hence increasing the potential for more asymmetric shocks in the future?

Section 2 provides results on the lack of the influence of external shocks on unemployment. Section 3 offers some new evidence on labour mobility in Europe, but otherwise relies on previous studies on migration in the US that do not confirm the perceived wisdom. Section 4 concludes.

2. Unemployment and asymmetric shocks

2.1 The optimum currency area approach: existing empirical literature

The standard line of reasoning in support of exchange-rate flexibility is the following: if a shock reduces the demand for the exports of a country, a real depreciation is required to maintain full employment and external equilibrium. The required real depreciation could also be achieved by a reduction in nominal ('money') wages, but this takes time and can presumably be achieved only if there is a period of substantial unemployment. The proper exchange-rate policy could thus reduce, and possibly even eliminate, the unemployment problems that arise from 'asymmetric shocks'. This line of reasoning has become the standard argument against EMU. Asymmetric shocks, it is often argued, will invariably ratchet up unemployment.

However, the available studies on the potential importance of this effect do not attempt to test this line of reasoning directly. They usually analyse the degree to which various macroeconomic indicators, e.g. output, the real exchange rate, unemployment, etc., are correlated across countries. A finding that these correlations are low is then usually interpreted as implying that the countries concerned are subject to important asymmetric shocks and that they would sustain large economic costs if they formed a monetary union.

What degree of correlation is acceptable is difficult to decide a priori since there is no theoretical reason to accept a correlation coefficient of, say 90 per cent, as sufficiently high for EMU, but reject anything below. This is why the implicit or explicit benchmark is often the US in the sense that it is argued that if the economies of member countries show a similar degree of correlation among them, as do states or regions inside the US, EMU should not create particular problems for Europe.

Many previous studies have followed this approach. It is sufficient here to take just one prominent example that can stand for most of this literature. Bayoumi and Eichengreen (1994) compare the correlation of certain shocks to output among 8 regions within the US and among 11 member states within the EU.(2) They distinguish between shocks that have transitory effects, which they assume to be demand shocks, and shocks that have permanent effects, which they assume to be supply shocks. Their main finding is that the supply shocks, thus defined, are larger in magnitude and less correlated across regions in Europe than in the US whereas the opposite holds for demand (i.e. transitory) shocks. Moreover, they also confirm that the core of the EU (here D, F, BE, NL and DK) constitutes a homogenous sub-unit. Within this restricted group of countries, supply (i.e. permanent) shocks are of roughly the same magnitude and cohesion as in the US. Their conclusion is that a core EMU is economically advisable, but not a wider EMU.

This example also shows one key problem of the empirical literature on the optimum currency area approach: the correlations in macroeconomic variables found for the past reflect not only the working of true shocks (i.e. 'intrinsic' factors like taste and technology), but also, and perhaps mainly, the extent to which monetary and fiscal policy have in the past tended to move together across countries (under different exchange-rate regimes).(3) Bayoumi and Eichengreen (1994) try to take this into account by distinguishing between supply shocks (presumably independent of policy) and demand shocks that might come from monetary and/or fiscal policy. However, neither they, nor other contributions, take into account that the optimal currency area is based on the need to adjust the real exchange rate in response to external shocks. No existing empirical analysis of EMU makes the distinction between external and domestic shocks; this is crucial, as argued below.

A different way to search for asymmetric shocks looks at differences in economic structures, e.g. differences in the shares of output accounted for by different industries or the product composition of exports. The underlying hypothesis here is that countries that have different economic structures are likely to experience asymmetric shocks. Gros (1996a) provides a number of indicators along this line and shows that they can give quite different results. This approach can in principle provide some information on likely sources of shocks, but it cannot provide evidence on the size of the asymmetric shocks one should expect in reality.

The available literature thus looks only at the potential for asymmetric shocks or measures co-movements in macroeconomic variables without showing how external shocks lead to unemployment. The basic question that has not yet been addressed in the literature is: are the 'classic' asymmetric shocks, i.e. shocks to export demand actually an important determinant of unemployment? A subsidiary question would concern the role of exchange-rate adjustments in containing unemployment generated by shocks to export. These are key questions for any evaluation of EMU because if the answer to both questions is yes (i.e. external shocks and the exchange rate are important for unemployment) one would have to conclude that the costs of EMU are high.

2.2 A direct test of the optimum currency area approach

How could one measure to what extent external shocks affect unemployment? In principle there are two ways: i) One could try to measure 'only' the extent to which (changes in) exports have 'caused' (changes in) employment or unemployment in the past. ii) one could use a large macroeconomic model which traces the impact of such shocks (e.g. changes in export demand) through the entire economy under various assumptions about the flexibility of wages and the exchange rate.

a) A direct link from exports to employment? Some results using the second method are reported further later in this section. This sub-section uses mainly the first method based on standard 'causality tests'. The underlying hypothesis in this case is that export supply is rather stable so that one can equate actual changes (innovations) in exports with changes in export demand. All the results presented here are based on a comparison of two regression equations:

a) [de.sub.t] = [Alpha] + [[Sigma].sub.i=1] [a.sub.i] [de.sub.t-i] + error term

b) [de.sub.t] = [Alpha] + [[Sigma].sub.i=1] [[Alpha].sub.i] + [[Sigma].sub.i=1] [[Beta].sub.i] [dexp.sub.t-i] + error term

Where [Sigma] stands for a summation that starts with the element indicated in the subscript, [de.sub.t] stands for the percentage change in employment (between period t and t-1) [TABULAR DATA FOR TABLE 1 OMITTED] and [dexp.sub.t-i] stands for the percentage change in export volumes between period t-i and period t-i-1. Exports (measured by various indicators as explained below) can then be said to 'cause' changes in employment if the [Beta]s, i.e. the coefficients on past and contemporaneous exports are together significantly different from zero. In other words, these tests measure the impact of (changes in various measures of) export performance on (changes in) employment once the autonomous movements in employment have been taken into account by including lagged employment changes among the explanatory variables. (One has to use percentage changes as the levels of both variables are clearly non-stationary.) A significant effect (of whatever sign) implies that one can reject at the 5 per cent confidence level the hypothesis that exports do not influence employment.

The first test involved looking at the link between industrial employment (index 1990=100, as reported by the IMF line 67 in the national country tables in International Financial Statistics) and export volumes (index 1990=100, also from the IMF) using data from 1960:1 to 1994:1.

Two equations were estimated for each of the ten countries. The first included the percentage change in employment as the dependent variable to be explained by a constant, four lags of the dependent variable and three quarterly dummies. The second equation added eight lags of the change in export volumes to the right-hand side.

The result was surprising: there is little evidence that shocks to exports influence industrial employment. Only five individual coefficients turned out to be significant at the conventional statistical level of 5 per cent.

Such a 'non' result is difficult to present. Table 1 summarises the regression statistics for both equations (but only for those five countries for which at least one coefficient on lagged export growth was significant). All the other countries did not even show a single significant coefficient on exports and an F-test did not lead to a rejection of the hypothesis that all coefficients were equal to zero. This result is all the more surprising since the industrial sector includes a relatively larger tradeable component. Thus, it would seem that the 'other factors' which affect changes in employment rates, beyond the autoregressive element coming from the past, are collectively far more important than shocks to exports.

The results in Table 1 are striking for two reasons. The first is that even for the countries selected for the greatest influence of exports the adjusted R-squared increases only marginally with the introduction of lagged changes in export volumes among the explanatory variables.

For those countries not included in this table (Germany, Denmark, Italy, Ireland and Portugal) the adjusted R-squared fell when the eight lags of export changes were introduced and so that the F-statistic should be below 1. Hence we do not bother to report the details of these regressions.

All in all these results suggest that for most member states growth in exports has not been a major factor in determining the evolution of employment (and unemployment) in the past. It is always very difficult to prove that a certain relationship does not exist, but given the similarity of the results using a number of different indicators of export demand this finding is rather strong. Moreover, there is little reason to believe that this will change in the future. Hence there is little reason to believe that shocks to the demand for exports will lead to significant unemployment problems in member countries under EMU.

The result that shocks to exports seem to have little statistically discernible impact on (un)employment could be interpreted in a number of ways. Apart from the testable objections discussed in the next section one interpretation could be that labour markets clear instantaneously. In this case EMU would not represent a problem. But this interpretation flies so manifestly in the face of all the other evidence on European labour markets (See OECD (1995) for example.)

b) Model simulations The approach followed so far has been totally a-structural. An alternative strategy would be to impose as much structure as possible by using a large model of the economy that allows one to calculate exactly the impact of a shock to export demand on output and other variables.

One example of this approach can be found in Emerson et al. (1990) who use a large econometric model of the EC (called Quest) which incorporates the short-run wage rigidity that is at the base of concerns about monetary union. Simulations with this model suggest that a 5 per cent shock to French export demand leads to a substantial fall in output (and prices) in France. If exchange rates are fixed French output falls by about 1.3 per cent in the first year and returns to baseline only by year seven. However, under flexible exchange rates the initial fall in output amounts still to 0.6 per cent and the subsequent recovery is actually slower so that the difference in present values of the GDP loss between fixed and flexible exchange rates is only 1.3 per cent. Recent simulations with the MultiMod model of the IMF communicated privately to the author show that the fall in output resulting from an exogenous fall in exports of 5 per cent is only one half of one percentage point of GDP higher under fixed exchange rates.

The optimal currency area approach (and this article) focuses on the impact of external shocks on employment and unemployment as opposed to output. However, unemployment and output are closely linked. For most countries the standard Okun curve type relationship translates a fall in GDP of I per cent into an increase in unemployment of about 0.3-0.5 per cent (in the short run). Emerson et al. (1990) also report that the standard deviation of export shocks is about 2.5 per cent. This implies that the difference between flexible and fixed exchange rates for a two standard deviation shock to exports would he only an increase of 0.2-0.3 percentage points in the unemployment rate (during the first year). External shocks would have to be unusually large under EMU to have a substantial impact on unemployment.

2.3 Types of shocks and the 'beggar-thy-neighbour' effect of the exchange rate

The results so far indicate that the standard shocks considered in the optimal currency area literature (i.e. shocks to export demand) do not have a major impact on the evolution of unemployment in Europe and that fixing exchange rates is not likely to make a large difference in this respect. Could one not argue that there are other types of shocks, which are empirically more important for unemployment, that could be better managed with flexible exchange rates? For example, it is often argued, that a major asymmetric demand shock like German unification did require an exchange-rate adjustment.

This episode is instructive because it can be used both as an argument for EMU and for greater flexibility of exchange rates. The standard argument is that after 1990 an overheating of domestic demand threatened price stability in Germany which forced the Bundesbank to adopt a very tight monetary policy. Since other EMS member countries, notably France, did not have the same problem, the optimal solution would have been to have lower interest rates in France than in Germany. This would have been compatible with a step appreciation of the D-mark followed by a gradual depreciation (i.e. greater exchange-rate flexibility). But one could also argue that if EMU had already existed the policy of the ECB would have been based on average area-wide inflation and its monetary policy would have been less restrictive.

However, in considering the use of the exchange rate for demand management purposes one has to keep in mind that an exchange-rate change shifts demand from one country to another. The real issue is thus the optimal exchange-rate policy from the point of view of the welfare of the system. This issue cannot be addressed with the usual one country models (which prescribe an exchange-rate adjustment in response to any internal shock, demand, supply or other). One has to use a two country model.

Gros and Lane (1994) use a standard two country model with short-run wage rigidity to analyse optimal exchange-rate policy in the presence of supply and demand shocks. They find that the Pareto optimum (which happens to coincide with the Nash equilibrium) is to let the exchange rate move in response to both shocks; but only if there are foreign shocks. This result implies that if two countries have a similar structure, so that shocks to the relative price of the goods they produce are unlikely, asymmetric shocks to domestic demand or supply are not a reason to keep exchange rates flexible. Different models might lead to slightly different results, but the basic intuition is likely to be robust to changes in the particular model used: from the point of view of the system there is no need to use exchange rates to distribute the impact of local shocks to demand if countries produce and consume the same goods.

This argument that at the global level the effects of exchange-rate changes on demand net out to zero does not apply to shocks that affect trade directly. If demand shifts from one country to another an exchange-rate adjustment is required from the point of view of both. Hence fluctuations in exports are the main source of shocks that should he taken into account to ascertain the importance of exchange-rate flexibility from a global point of view. Other legitimate sources of shocks would be external shocks (like an oil price change) that have differential effects because of differences in the importance of energy.

By contrast, one could imagine the case of a country which experiences a sudden fall in domestic demand because households suddenly save more. A depreciation would shift demand towards domestic goods and increase exports, thus reducing the unemployment that would otherwise result from the drop in demand. However, the 'gain' in demand of the country experiencing the shock would come at the expense of the rest of the world. The country that depreciates would only export its unemployment problems. From a global point of view little would be gained from exchange-rate flexibility in this case.

It is difficult to imagine in concrete terms economy-wide shocks that are driven by sudden changes in technology or tastes. While there might be sudden changes at the sectoral level experience indicates that these fundamental determinants of the economy tend to change slowly at the aggregate level, which should give prices and wages enough time to adjust to maintain equilibrium. For example, the rise in the importance of the car industry or the decline of railways took decades. These secular changes certainly caused severe adjustment problems, but the argument that adjustments in the real exchange rate can be achieved quicker through changes in the nominal exchange rate loses its significance for trends that work over a decade or more.

2.4 Why care about exchange rates?

The results so far raise the question: Why should politicians want EMU? If trade has little impact on unemployment politicians should not care either way. But this argument might be besides the point because exchange-rate variability, as opposed to the level of the exchange rate might still have an impact of the performance of the economy at home. In particular governments are (and will continue to be) held responsible for the state of the labour market which can be decisive for elections. The position important pressure groups, such as trade unions, take concerning EMU will also depend more on the labour market implications of exchange-rate variability than on its effect on the volume of trade.

Support for EMU should thus depend on the (perceived) impact exchange-rate variability has on employment and unemployment. Given the results found here one would assume a priori that it should be minor. But this presumption might be wrong, at least for a key country, namely Germany. The result of a simple causality type analysis reported in Box 1 shows that exchange-rate variability does have a significant impact on job creation in Germany. No similar effect was found for the level of the D-mark exchange rate.

This result shows that exchange-rate variability has a significant impact on employment in Germany.

The results of this section suggest thus not only the cost of abandoning the exchange-rate instrument have been over-rated, but also that eliminating exchange-rate variability could have substantial positive effects on its own.

3. EMU and labour mobility

The introduction mentioned the important place accorded to labour mobility in the optimal currency area approach. However, this view looks only at labour mobility as a short-run adjustment mechanism and does not take into account that concentration of industry and hence pronounced core-periphery patterns are more likely to emerge when labour mobility is high. But since most studies concur that labour mobility is low in Europe (not only across countries, but also across regions within countries (see Decressin and Fatas (1995) who again make comparisons with the us) there should be less concentration in Europe than in the US.

Some authors have used this line of thought to arrive at a sort of catch 22: as long as labour mobility is low in Europe EMU is costly because labour mobility is needed to offset asymmetric shocks. However, so the argument goes; if labour mobility were to increase (possibly because EMU comes anyway) concentration would increase and hence the likelihood of asymmetric shocks would also increase, again making EMU costly. The suggested conclusion is that heads EMU is impossible and tails it is not desirable. The proper conclusion would seem to be that labour mobility is perhaps less crucial for EMU than previously thought: although labour mobility allows for a quicker adjustment to shocks it also favours concentration of industry and hence increases the potential for asymmetric shocks.

However, since labour mobility is usually assumed to be important it is still useful to take a look at the data which do not always yield the results that are commonly expected.

3.1. Inter-national versus inter-regional mobility

It is a commonly accepted proposition that labour mobility in Europe is very low in absolute terms and compared to the US. A corollary is that the potential costs of EMU should be high. However, this corollary is not warranted because, as argued above the key consideration for the optimal currency area is the difference between inter-regional labour mobility within countries and labour mobility across countries. Both sides have so far not been documented systematically because of the absence of reliable statistical material. This is now changing,(4) and the data now available do not confirm the widely held notion of relatively low international labour mobility.

In 1992 almost 2.2 million immigrants came to the member states of the EU (equivalent to about 0.7 per cent of population). And it appears (these data are less reliable) that emigration was more than 1 million lower than immigration. This can be compared to the US where the average net immigration was about 800 thousand on average per annum during 1986-91 (about 0.4 per cent of population), lower than that of the EU in 1992.

If one wants to judge whether the observed level of migration in the EU indicates a degree of labour mobility that is so low that asymmetric shocks in an EMU will lead to serious problems, inter-regional migration within member states provides a useful reference point.(5) Table 3 therefore shows the most recent available data on immigration from the rest of the world as a percentage of the overall population and the percentage of the population that moved between regions within the country. Given that the data on emigration are much more partial only the data on immigration will be discussed below.(6)

Table 3. Migration in Europe

 % of population In thousands

 International
 International migration Inter- Inter-
 (immigration (within national regional
 into member member
 countries) countries)

From EU15 0.67 0.89 2 356 3 313 -

Source: Eurostat

N.B.: Immigration into SW, SF, L IRL, GR, DK (all the countries for
which no data on inter-regional migration is available) was equal
to
186.9 thousands.

Table 3 shows that the total number of immigrants arriving in EU countries, about 2.3 million, is below the number of inter-regional migrants, about 3.1 million. However, the orders of magnitude are similar. International migration amounts to more than 2/3 of inter-regional migration.

Hence it appears that contrary to what economists have so far assumed inter-regional and inter-national migration are of a similar order of magnitude in Europe. What conclusions can one draw from this? While one should not presume automatically that inter-regional migration within member states is sufficient to make them optimum currency areas one can at least conclude that - for a given importance of asymmetric shocks - a monetary union for the EU should not create more problems than the monetary unions coinciding with existing nation states create at the regional level, provided one can assume that international migrants are flexible in the choice of their country of destination.

3.2 The contribution of labour mobility to adjustment

It is apparent that people move much more often in the US than in Europe. However, what matters in the context of discussions about EMU is the extent to which net movements react to local unemployment. It is surprising to note how little hard evidence exists on this point. The most widely cited study is by Eichengreen (1993) who compares the reaction of inter-regional migration to local unemployment and wages in the US, UK and Italy. He finds that net immigration to any of the 9 US Census regions indeed reacts to unemployment in the previous period, however, the effect is rather imprecisely estimated since the t-statistic is only 1.92.(7) The point estimate (-0.37) implies that net immigration would fall only by 0.0825 (percentage points) if the average unemployment for the US is 8 per cent and if it increases in any region from this level to 10 per cent. If migrants have the same family composition and activity rates as the local population the change in migration would thus be equivalent to 1/25th of the increase in unemployment.(8)

Blanchard and Katz (1992) report a much stronger reaction of migration to unemployment. They estimate that a negative shock to employment in any 'average' US state is offset within one period by about 60 per cent through migration. Their finding runs counter to many other studies on the US labour market, which generally find, as reported in Greenwood (1975) and (1985) that unemployment is not an important factor in explaining migration flows. This discrepancy might be due to the fact that Blanchard and Katz do not use any direct data on migration, but calculate migration as a residual from data on the labour force, employment and unemployment. Since these data come from different sources it is likely that some of their coefficients pick up inconsistencies in the data (i.e. a measurement error) which are strongly correlated with the other variables. Since migration is really the residual the estimated effects of an unemployment shock are not based only on the migration that actually takes place but also on the inconsistencies in the data.(9)

3.3 Reasons for low labour mobility in Europe

It is often argued that international labour mobility is (and will remain) low in Europe because of cultural and social barriers. However, why should inter-regional labour mobility be also rather low? One reason might be the labour market and the fact that once income exceeds a certain threshold people are no longer willing to incur the psychological cost of moving (see Faini, 1994). But this cannot be a full explanation because there is also considerable variation among member states in the rate of domestic migration between regions.

One factor that is often overlooked is the housing market. The most important pecuniary (and perhaps also psychological) cost of moving for most people is that it involves a change of housing. In countries where this market is not flexible this factor might be decisive as shown by the evidence in Box 2. Making the housing market more flexible could thus be as important as a reform of the labour market in preparing for EMU.

This brief analysis of the link between labour mobility and the housing market can only be suggestive. It confirms that the structure of the housing market does certainly have an influence on labour mobility. This result is not new (see for example Hughes and McCormick (1987)), but the issue has never been approached systematically on a cross-country basis.

The results reported here suggest that owners of housing are less mobile than others. A proper analysis would have to establish that owning a house is not just a proxy for another characteristic that pre-disposes the household for higher inter-regional mobility. This is less likely to be a problem in a cross-country analysis than in studies on national labour markets. To address this problem at a European level would require considerable additional work. The work of Hughes and McCormick (1987) suggests that this is perhaps not a crucial problem. They report that tenants of council housing (at the time about one third of the UK population) moved quite often within their region, but almost never outside their region. Their observation implied that if local council housing were to be sold inter-regional mobility could substantially increase in the UK. This seems indeed to have happened since the large scale privatisation of council housing. Further cross-country research in this direction might add an important new angle to the optimal currency area argument.

4. Conclusions

The main empirical finding of this article is that employment in the past was not influenced in a significant way by changes to export demand. Hence it is not likely that the lack of exchange-rate adjustments under EMU will lead to major problems in this area. The standard argument that EMU will lead to less employment because external shocks could no longer be offset through exchange-rate changes has been exaggerated.

International labour movements in the EU (especially immigration from third countries) have now increased to a point were they are of a comparable order of magnitude as inter-regional migration within member countries. EMU should thus not be more difficult to manage than the existing monetary unions in Europe that member states represent. Reducing barriers to labour mobility remains, of course, desirable at any rate. Making the housing market more flexible could contribute considerably to this goal.

Box 1. Exchange-rate variability and the German labour market

This box reports the results of some simple causality tests for the influence of the variability of the DM exchange rate (against the currencies of the 7 other original members of the ERM: B-LUX, DK, F, IRL, IT, NL) on employment growth. Only European exchange rates are used because only their variability could be suppressed by EMU. These countries represent also the most likely early candidates for membership in EMU. The initial ERM candidates were used because when the EMS was created politicians used to emphasise the gains from exchange-rate stability.

The result reported below are again standard causality tests on annual data. The exchange-rate variability of the DM was measured by taking for each year the standard deviation of the 12 month-to-month changes in the logarithm of the nominal exchange rate of the DM against the currencies of the countries mentioned above. These 7 standard deviations were then aggregated in one composite measure of exchange-rate variability (denoted by 'exv' below) weighting them by the weights of the countries in the ECU (which correspond approximately to their weights in terms of GDP).

Performing the same analysis for the other major European countries revealed that France and Spain showed a similar pattern as Germany. The growth rate of employment is systematically related to past changes in the intra-ERM variability of the respective national currencies. No similar pattern was found for the UK and for the US. In the case of the US two variability measures were used: i) the volatility of the dollar against European currencies, and ii) the volatility of the effective exchange rate of the dollar. In both cases the result was negative in the sense that once was not forced to reject the null hypothesis that exchange-rate variability had no impact on employment creation. This is in a certain sense reassuring since one would not expect that a large continental economy like the US is influenced by exchange-rate variability.

Table 2. Exchange rate variability and the German labour market

Explanatory variables: Dependent variable:
 percentage change in employment

 Coefficient (t-statistic)

Constant 1.66 (5.7)

Lag 1 0.79 (5.6)

lag 2 -0.54 (4.5)

Exc(-1) -1.3 (5.3)

 Adjusted R-squared 0.78
 Mean of dependent var 1.34
 S.E. of regression 0.63
 Durbin-Watson 2.05
 F. statistic 29.21687

Source: Gros (1996b)

The results of this section suggest thus not only that the cost of abandoning the exchange-rate instrument has been over-rated, but also that eliminating exchange-rate variability could have substantial positive effects on its own. The first tests reported here can, of course, only be suggestive. The main problem, that cannot ever be fully resolved, is that exchange-rate variability could just be a proxy of uncertainty elsewhere in the economy. However, it has not so far been possible to identify 'causes' of exchange-rate volatility. In particular exchange-rate variability is apparently not systematically linked to the volatility of monetary policy (see Rose (1997). For the case of Germany a number of potential alternative explanations were therefore tested by Gros (1996b). However, the hypothesis considered there (e.g. that exchange-rate variability increases because the Bundesbank increases interest rates) did not affect the result reported here. A similar sensitivity analysis, with a wider array of alternative explanation would need to be performed for other countries as well.

This section has not established beyond doubt that exchange-rate variability has a strong impact on labour markets. But the results reported here do cast some doubt on the presumption of many economists that exchange-rate variability is not important on its own because most studies have shown that it has little influence on the volume of trade.

Box 2. Migration and the housing market

Figure 1 shows that there are large differences among member countries in the rate of inter-regional migration which ranges from 0.4 per cent in Spain to almost 1.9 per cent in the UK. What could cause large differences in the rate of inter-regional migration? The size of regions as defined by Eurostat is similar across countries and should not be a decisive factor.

Differences in unemployment rates and wages should constitute the main incentive for migration. The main obstacle (at least within one country) to mobility should be rigidities in the housing and labour markets. The more flexible they are the easier it should be for people to move. However, these two are the most heavily regulated markets in most member states. The overall degree of flexibility of the labour market is impossible to measure. However, for the housing market the possibility exists by using data on the proportion of households that own their homes. These data are not easily available. However, financial markets associations publish data on the proportion of houses that are occupied by their owners. The higher this proportion the more difficult it will be for people to move because it is usually much easier and cheaper to change between rented accommodations than to sell a home and buy another one.

Figure 1 below shows that on the continent there is indeed a strong correlation between the rate of inter-regional migration and the proportion of houses occupied by their owners in 1991 or 92 (see EMF, (199)). The more people own the place they live in the less significant is labour mobility. This figure also suggests that the UK constitutes an outlier in the sense that there is more inter-regional movement in the UK than one would expect given the rather high rate of owner occupancy. Since there are only 6 large countries for which it makes sense to speak of inter-regional migration there are not enough degrees of freedom for a real statistical analysis. The regression line without the UK shown in Figure 1 is thus only suggestive.

One possible explanation for the higher rates of migration and lower dispersion of unemployment in the UK might be that the reforms under Thatcher have made the labour market much more flexible. Until only 5 years ago the dispersion in regional unemployment was much higher in the UK (data on migration for the past are more difficult to obtain). Moreover, the housing market is much more flexible in the UK than on the continent where rules on the proportion of the value of the house that can be obtained through a mortgage are much tighter and where the taxes on the sale of houses are typically about 20 per cent of the value of the house sold.

The authors are grateful to the Editorial Board for extremely useful comments on an outline draft of this article and to Oxford Economic Forecasting for help with data as well as continuing discussions of the issues. Especial thanks are due to Alison Gomm for invaluable aid during the production process. The usual disclaimers, of course, apply.

NOTES

(1) The regional dimension is often overlooked in discussions about EMU because it has to be assumed that the alternative to EMU is the continuing existence of national currencies, and not the introduction of regional currencies. However, most European regions are of a similar size as the average state of the US which are often compared to member countries.

(2) A somewhat different approach can be found in De Grauwe and Vanhaverbeke (1993) who analyse the variability of real exchange rates across regions and countries. The finding that real exchange rates vary significantly more across countries than across regions within a country is difficult to interpret: Is it due to an excess volatility of exchange rates or are there large asymmetric shocks (policy or other) that provoke this exchange-rate variability?

(3) One could thus argue that the high correlations found for the core countries are probably an underestimate of the correlations that would result under EMU (i.e. with a unified monetary policy). It can also not be excluded that some of the countries that had lower correlations in the past would actually belong to the core once they also belong to EMU.

(4) See 'Statistics in Focus' 1995, 3 of Eurostat that concentrates on international migration concerning EU member states.

(5) The most recent available data on this is for 1990-2, but it appears that inter-regional migration within member states has been rather stable over the last decade.

(6) For inter-regional movements within member states emigrants equal immigrants by assumption. One should also keep in mind that the national definitions of what constitutes an immigrant (or migrant) vary greatly.

(7) However, the constant term in his equation is rather precisely estimated (t-statistic of 5.76) and indicates that immigration amounts each year to about 1.1 per cent of the population of the region if the region has the same wage rate and unemployment rate as the average for the entire US. The constant term is about 10 times higher for the US than for the UK.

(8) Bayoumi and Prasad (1995) analyse the behaviour of sectoral employment in some member states and US regions. They find that most of the shocks to employment are industry specific and in the US and European countries, but they chose to interpret the same result differently: for the US this result is taken to indicate a high degree of labour mobility because wages are also mostly affected by industry specific shocks whereas in Europe this result is taken as an indication of low labour mobility because shocks to wages are mostly country specific.

(9) However, Blanchard and Katz's approach seems to have internal difficulties which become apparent once one applies the same methodology to a European country. As an illustrative example I replicated their methodology using data from Germany (relative to the EU) average. The tri-variate system (percentage change in employment, employment rate and participation rate, all Germany relative to us) seems to work well. It yields a sort of 'Okun coefficient' of about 0.33, similar to the one found by them for the average US state. Moreover, and this is crucial, the reaction of the labour force participation rate to employment shocks is larger than in the US, but still moderate: a fall in employment of 1 per cent leads to a fall in participation of 0.39 per cent in the first year. The 'implicit' migration would thus be for Germany 0.28 (the result of 1 - 0.33 - 0.39). This is still too high to be believable. The dynamics of the system implies that after 3 years, the loss of employment has increased to 2 per cent, unemployment has increased only to 0.42 per cent (its peak) and labour force participation has fallen by a cumulative 1.2 per cent, meaning that 38 per cent of the number initially fired should have emigrated in the meantime.

REFERENCES

Bayoumi, T. and Eichengreen, B. (1994), 'Shocking aspects of European monetary integration', in F. Giavazzi and F. Torres (eds) Adjustment and Growth in the European Monetary Union, Centre for Economic Policy Research (CEPR), Cambridge University Press.

Bayoumi, T. and Prasad. E. (1995), 'Currency unions, economic fluctuations and adjustment: some empirical evidence', CEPR Discussion Paper no. 1172, May.

Blanchard, O. and Katz, L. (1992), 'Regional evolutions', Brookings Papers on Economic Activity I: 1992.

Decressin, J. and Fatas, A. (1994), 'Regional labour market dynamics in Europe', CEPR Discussion Paper no. 1085.

De Grauwe, P. and Vanhaverbeke, W. (1993), 'Is Europe an optimum currency area?: evidence from regional data', chapter 5 in Policy Issues in the Operation of Currency Unions, edited by Paul R. Masson and Mark P. Taylor, Cambridge University Press.

Eichengreen, B. (1993), 'Labour markets and European monetary unification', chapter 6 in Policy Issues in the Operation of Currency Unions, edited by Paul R. Masson and Mark P. Taylor, Cambridge University Press.

Emerson, M. et al. (1990), 'One market, one money', European Economy, no. 44, October.

EMF (199), Annual Report, Brussels.

Eurostat (1995), 'Statistics in focus', Populations and Social Conditions, no. 3.

Faini, R. (1994), 'Trade unions, fiscal policy and regional development', The Optimal Currency Area of Economic Activity: New Theories and Evidence, CEPR Conference Report.

Greenwood, M.J. (1975), 'Research on internal migration in the United States: a survey', Journal of Economic Literature, vol. 13.

Greenwood, M. J. (1985) 'Human migration: theory, models and empirical studies', Journal of Regional Science vol. 25.

Gros, D. and Lane, T. (1994), 'Symmetry versus asymmetry in a fixed exchange rate system', Kredit und Kapital, Heft 1, 1994.

Gros, D. (1996a), 'Towards Economic and Monetary Union: problems and prospects', Centre for European Policy Studies Paper no. 65.

Gros, D. (1996b), 'Germany's stake in exchange rate stability', Centre for European Policy Studies, Centre for European Policy. Studies Review, vol. 1, May.

Hughes, G. and McCormick, B. (1987) 'Housing markets, unemployment and labour market flexibility in the UK', European Economic Review, no. 31.

Mundell, R. (1961), 'A theory of optimum currency areas', American Economic Review, vol. 51, September.

OECD (1995), 'The OECD jobs study', OECD, Paris.