Chapter

III Long-Term Costs of Losing Monetary Policy

Author(s):
International Monetary Fund
Published Date:
April 2005
Share
  • ShareShare
Show Summary Details

The principal economic costs of joining a currency union stem from the effects of giving up an independent monetary policy on the stability of growth and inflation. A long tradition of assessing these costs exists in the optimum currency area (OCA) literature, which began as an analysis of the characteristics of countries that would be well suited to forming a currency union but has evolved into an examination of the effectiveness of monetary policy.

The early OCA analysis focuses on criteria for defining optimum currency areas in terms of two basic attributes: susceptibility to asymmetric shocks and the efficiency of mechanisms other than monetary policy for adjusting to shocks (Mundell, 1961; Kenen, 1969). The pioneering studies, done in an era of limited capital mobility, focus on countries whose main interactions are through trade. They conclude that a country is better suited to currency union (1) the more similar its structure of production is to that of the union and the greater is its product diversification—attributes that reduce a country’s susceptibility to asymmetric shocks; and (2) the more flexible are its wages and prices and the more mobile is its labor force—attributes that determine a country’s capacity for adjusting to asymmetric shocks in the absence of exchange rate changes. The value of an independent monetary policy is also viewed as less important for countries highly open to international trade. In these circumstances, the effects of changes in monetary policy are more likely to be dampened by demand spillovers to trade, and, with a small nontradables sector, exchange rate changes are more likely to result in price rises evoking a strong wage reaction (McKinnon, 1963). Subsequent scrutiny of these simple conclusions has left them largely intact but attenuated in some important respects.

Judgments on susceptibility to shocks are influenced by two new considerations. First, the scope afforded for international portfolio diversification by the increase in capital flows over the past few decades permits households and firms to self-insure against country-specific shocks. This weakens the importance of synchronization of shocks as a prerequisite for monetary union. Second, OCA characteristics are increasingly seen not as static, but as subject to ex post or endogenous convergence after joining a currency union: that is, a country’s susceptibility to asymmetric shocks may change rapidly once it joins a currency union. Specifically, by increasing trade links, a currency union may change the structure of a country’s trade so as to make countries more or less susceptible to asymmetric shocks—the former if it leads to greater specialization of production and more interindustry trade (Krugman, 1991c; Artis, 1991) and the latter if it leads to more intraindustry trade (Frankel and Rose, 1996). Available empirical evidence on which of these effects will dominate is tilted toward greater intraindustry trade and rising correlation of shocks.

Similarly, mechanisms besides monetary policy for adapting to asymmetric shocks may change after joining a currency union. A country may become a better candidate for a currency union after entry if loss of the exchange rate instrument stimulates adjustment through other channels such as wage and price flexibility (Calmfors and Johansson, 2002). Buiter (2000) in an application of the Lucas critique suggests that the degree of nominal rigidities is the outcome of purposeful choices, and joining a currency union could well reduce nominal rigidities.

Views on monetary policy have also changed profoundly since the early OCA literature, generally weakening the case for an independent monetary policy. Certainly, giving up monetary policy when fiscal and structural policies are not in line with those of the currency union is unambiguously untenable. But once they are compatible, the gains from an independent monetary policy are less clear. From a long-term perspective, it is now widely agreed that monetary policy cannot exercise any trade-off between inflation and growth—long-run Phillips curves are vertical. In fact, in the long run, a credible, imported monetary policy may even reduce the burden of achieving and maintaining low inflation. From a short-term perspective, however, monetary policy is still seen as effective in countering some types of temporary shocks, although only in the presence of nominal rigidities and factor immobility.

Most evidence suggests that exchange rate changes are better suited to countering demand shocks than supply shocks (although this depends, inter alia, on whether the authorities’ objective function is focused on smoothing output, absorption, inflation, or balance of payments fluctuations). But fixing the exchange rate is optimal in the face of financial shocks so as to avoid transmitting nominal shocks into relative price changes (Calvo and Mishkin, 2003). Indeed, when asset market developments dominate, exchange rate movements will reflect not only fundamentals but also collective mood swings and noise trading (Flood and Rose, 1995; Jeanne and Rose, 1999). Thus, the benefits of exchange rate flexibility for absorbing shocks are bundled with the costs of possibly volatile and unwarranted movements (Buiter, 2000).

The upshot of these considerations is that a small, open economy that has fiscal and structural policies compatible with a currency union may lose little by giving up its exchange rate instrument, while gaining monetary policy credibility and protection from financial market shocks.

This chapter examines how valuable independent monetary policies are for the CECs. Three prongs of analysis are pursued: countries’ performance on OCA criteria are assessed; a vector autoregression (VAR) model identifying the nature of shocks and effectiveness of the exchange rate in addressing them is estimated; and simulations of the Taylor efficiency frontier (TEF) using the IMF’s Global Economic Model (GEM) for a country with characteristics representative of a CEC outside and inside currency union are compared.

How Do the CECs Stack Up on the OCA Criteria?

Notwithstanding the many qualifications, susceptibility and adaptability to asymmetric shocks remain the gold standards for judging the appropriateness of membership in a currency union. Both are likely to be endogenous to currency union membership so that static, backward-looking measures of them are imperfect for a forward-looking assessment. But without a consensus on how currency union membership will affect them, recent history remains the best gauge.

Susceptibility to Asymmetric Shocks

Three types of measures of susceptibility to asymmetric shocks exist: direct correlation of shocks, correlations of growth rates of economic activity so as to capture both asymmetric shocks and asymmetric responses to symmetric shocks, and measures of structural features of economies that make them more or less prone to asymmetric shocks or asymmetric responses to shocks.

Recent empirical evidence points to low, but probably rising, correlations of shocks between the CECs and the euro area. Two studies (Fidrmuc and Korhonen, 2002; Frenkel and Nickel, 2002) examine quarterly data from overlapping periods since 1991 in VAR models that distinguish between supply and demand shocks (Table 3.1).9 Their results suggest a few broad conclusions. First, correlations of shocks between the CECs and the euro area as a whole vary markedly but, for the most part, are low relative even to the noncore euro-area countries. Second, for all countries, but especially for the CECs, correlations of supply shocks are substantially higher than correlations of demand shocks. This is somewhat reassuring insofar as correlations of demand shocks are probably heavily influenced by the degree of harmonization of monetary and fiscal policies, while correlations of supply shocks probably represent more fundamental exogenous influences. Third, correlations of supply shocks appear to be rising over time; for three of the five CECs, supply shock correlations are markedly higher in Frenkel and Nickel’s study (which uses data from a more recent period) than in Fidrmuc and Korhonen, while correlations of demand shocks are lower. This suggests that some convergence of influences on long-term growth is occurring, even if divergent short-term influences (possibly related to macroeconomic policies) are increasing.

Table 3.1.CECs and Euro-Area Countries: Correlations of Shocks with the Euro Area1(Coefficients of correlation)
Fidrmuc and Korhonen (2002)Frenkel and Nickel (2002)
1991–200021995–20013
SupplyDemandSupplyDemand
CECs
Average0.160.030.24–0.10
Minimum0.04–0.18–0.69–0.43
Maximum0.460.280.730.22
Czech Republic0.04–0.150.34*–0.24
Hungary0.46*0.25*0.73*0.12*
Poland0.080.28*–0.690.22*
Slovak Republic0.05–0.050.18*–0.43
Slovenia0.15–0.180.66*–0.15
Core euro
Average0.550.120.510.20
Minimum0.380.00–0.01–0.58
Maximum0.690.301.000.94
Austria0.380.080.18–0.04
Belgium0.530.001.000.94
France0.690.300.740.35
Germany0.660.180.620.31
Netherlands0.470.04–0.01–0.58
Noncore euro
Average0.230.170.280.12
Minimum–0.14–0.01–0.18–0.10
Maximum0.520.570.760.55
Finland0.300.060.54–0.10
Greece0.05–0.01–0.10–0.01
Ireland–0.140.13
Italy0.520.570.760.55
Spain0.220.160.420.03
Portugal0.450.09–0.180.11
Average for euro area0.380.150.400.16
Sources: Fidrmuc and Korhonen (2002); and Frenkel and Nickel (2002).

Indicates correlation estimates that are positive and above the minimum of the euro core.

Correlation coefficients of shocks in individual countries with the aggregate of the euro area, based on VAR estimates.

Quarterly data, 1991–2000 for the euro-area countries; 1993–2000 for the CECs, except Hungary; 1995–2000 for Hungary.

Quarterly data, 1995–2001.

Sources: Fidrmuc and Korhonen (2002); and Frenkel and Nickel (2002).

Indicates correlation estimates that are positive and above the minimum of the euro core.

Correlation coefficients of shocks in individual countries with the aggregate of the euro area, based on VAR estimates.

Quarterly data, 1991–2000 for the euro-area countries; 1993–2000 for the CECs, except Hungary; 1995–2000 for Hungary.

Quarterly data, 1995–2001.

Correlations of GDP and industrial production growth rates between individual countries and Germany reinforce these conclusions (Table 3.2). These correlations, which capture the symmetry of both shocks and responses in individual countries to symmetric or asymmetric shocks, also vary over quite wide ranges. But at least in Hungary and Slovenia, they are as high as in several of the core and noncore euro-area countries. Moreover, these results bear out the earlier observation that correlations in all the CECs are rising. Presumably this reflects the diminution of transition-related shocks while economic integration with the euro area has deepened via trade, FDI, and financial market integration.

Table 3.2.CECs and Euro-Area Countries: Correlations of Indicators of Economic Activity with Germany(Coefficients of correlation)
Period covered VariableFidrmuc (2001)Boreiko (2002)2IMF
1993–9911993–20011997–20011999–20011996–2002
Industrial productionGDPIndustrial productionIndustrial production3,4
CECs
Average0.450.540.380.530.640.510.43
Minimum0.040.010.200.450.520.330.24
Maximum0.770.800.520.600.840.790.62
Czech Republic0.37*0.010.200.500.520.400.31
Hungary0.63*0.75*0.520.600.840.79*0.62*
Poland0.45*0.380.390.590.680.500.44*
Slovak Republic0.040.74*0.300.530.580.520.24
Slovenia0.77*0.80*0.490.450.560.330.54*
Core euro
Average0.630.750.760.61
Minimum0.250.580.630.41
Maximum0.910.880.890.77
Austria0.810.580.890.77
Belgium0.250.880.630.50
France0.910.830.860.77
Netherlands0.570.690.680.41
Noncore euro
Average0.660.630.720.53
Minimum0.48–0.030.380.11
Maximum0.920.810.920.73
Finland0.690.790.920.73
Greece0.480.600.51
Ireland–0.030.800.52
Italy0.810.850.70
Spain0.920.790.780.61
Portugal0.560.780.380.11
Average for euro area50.650.680.740.56
Sources: Fidrmuc (2001); Boreiko (2002); IMF, International Financial Statistics; Eurostat; and IMF staff estimates.

Indicates correlation estimates that are above the minimum of the euro core.

Year-on-year growth rates of quarterly indicators of economic activity.

Monthly industrial production growth rates.

Year-on-year growth rates of monthly industrial production.

Monthly industrial production indices detrended using a Hodrick-Prescott (HP) filter.

Excludes Germany and Luxembourg.

Sources: Fidrmuc (2001); Boreiko (2002); IMF, International Financial Statistics; Eurostat; and IMF staff estimates.

Indicates correlation estimates that are above the minimum of the euro core.

Year-on-year growth rates of quarterly indicators of economic activity.

Monthly industrial production growth rates.

Year-on-year growth rates of monthly industrial production.

Monthly industrial production indices detrended using a Hodrick-Prescott (HP) filter.

Excludes Germany and Luxembourg.

Structural characteristics of the CEC economies are also changing in ways that should reduce the incidence of asymmetric shocks. In particular, for each country, the share of intraindustry trade (measured at the two-digit Standard International Trade Classification (SITC) level) in total manufacturing trade has risen sharply since the early 1990s. As of 1996–2000, indices stood at levels comparable to those in the euro-area core and well above the non-core (Table 3.3).10 Although sectoral distributions of employment and GDP continue to differ quite widely between the CECs and the euro area (industry being substantially higher and services lower), structures of manufacturing sectors look quite similar: measures of sectoral asymmetries in manufacturing in the CECs relative to those in the euro area are well inside the euro-area extremes (Tables 3.4 and 3.5).

Table 3.3.CECs and Euro-Area Countries: Manufacturing Intraindustry Trade1(In percent of total manufacturing trade)
1988–911992–951996–2000
CECs
Average55.766.072.5
Minimum54.961.762.6
Maximum56.469.877.4
Czech Republic66.377.4
Hungary54.964.372.1
Poland56.461.762.6
Slovak Republic69.876.0
Slovenia267.874.6
Core euro
Average72.374.472.8
Minimum67.170.468.9
Maximum77.677.777.5
Austria71.874.374.2
Belgium377.677.771.4
France75.977.677.5
Germany67.172.072.0
Netherlands69.270.468.9
Noncore euro
Average56.257.157.1
Minimum42.839.536.9
Maximum68.272.171.2
Finland53.853.253.9
Greece42.839.536.9
Ireland58.657.254.6
Italy61.664.064.7
Spain68.272.171.2
Portugal52.456.361.3
Average for euro area63.564.964.2
Source: OECD.

Intraindustry trade is calculated with the conventional Grubel-Lloyd (G-L) index, using two-digit SITC (revision 3) product classes. For any product i, the share of intraindustry trade in the product class i between countries A and B is given by the following ratio: IITI,AB = 1 – (|XiMi|)/(Xi + Mi). The G-L indices are weighted averages of these indices for all products of class i.

Calculated by the Slovenian authorities.

Belgium and Luxembourg.

Source: OECD.

Intraindustry trade is calculated with the conventional Grubel-Lloyd (G-L) index, using two-digit SITC (revision 3) product classes. For any product i, the share of intraindustry trade in the product class i between countries A and B is given by the following ratio: IITI,AB = 1 – (|XiMi|)/(Xi + Mi). The G-L indices are weighted averages of these indices for all products of class i.

Calculated by the Slovenian authorities.

Belgium and Luxembourg.

Table 3.4.CECs: Sectoral Distribution of Employment and GDP, 2001(In percent of total)
EmploymentGDP
AgricultureIndustryServicesAgricultureIndustryServices
Czech Republic5415544056
Hungary6355943264
Poland19315033364
Slovak Republic6375753363
Slovenia10395133760
CEC average9365443561
Euro area4266922771
Source: Eurostat.
Source: Eurostat.
Table 3.5.CECs and Euro-Area Countries: Manufacturing Asymmetry Indicator1
Ireland102.1
Finland70.3
Netherlands55.1
Portugal52.2
Hungary51.8
Greece49.8
France43.1
Slovak Republic40.6
Austria38.1
Poland38.1
Czech Republic37.0
Italy28.7
Spain26.2
Slovenia25.32
Germany21.5
Sources: OCED Industrial Surveys; and Slovenian authorities.

Sum of absolute values of the differences between the shares of gross manufacturing output by industry and euro-area shares. The higher the indicator, the greater the deviation from the average manufacturing structure of the euro area. Data for 2000 (except 1999 for the Czech Republic, Greece, Ireland, Italy, and 1997 for the Slovak Republic).

Calculated by Slovenian authorities.

Sources: OCED Industrial Surveys; and Slovenian authorities.

Sum of absolute values of the differences between the shares of gross manufacturing output by industry and euro-area shares. The higher the indicator, the greater the deviation from the average manufacturing structure of the euro area. Data for 2000 (except 1999 for the Czech Republic, Greece, Ireland, Italy, and 1997 for the Slovak Republic).

Calculated by Slovenian authorities.

However, foreign portfolio diversification by the CECs—which allows self-insurance against asymmetric shocks—remains a fraction of that in representative euro-area countries. The ratio of foreign assets to GDP ranges from 25 percent in Poland to 68 percent in the Czech Republic, compared with well over 100 percent in selected euro-area countries (Table 3.6). In part this difference reflects the relatively low savings rates in CECs, the recent opening of capital markets, and some lingering restrictions on pension fund holdings of foreign assets. Rapid change in this form of self-insurance should be expected after euro adoption.

Table 3.6.CECs and Selected Euro-Area Countries: International Investment Positions (IIPs), 2002(In percent of GDP)
Czech RepublicHungaryPolandSlovak RepublicSloveniaGermanyItalyPortugalSpain
International Investment Positions, net1–19.3–57.2–36.0–21.8–6.110.4–2.8–45.724.6
Assets68.236.825.454.757.1145.491.8141.7102.5
Liabilities87.494.061.476.563.2135.094.7187.5127.1
Direct investment52.242.323.831.817.123.19.932.429.9
Portfolio investment28.829.511.812.59.132.350.754.544.5
Other investment25.422.225.832.237.079.734.0100.552.7
Sources: National central bank websites; IMF, World Economic Outlook; and IMF staff estimates.

Foreign assets minus foreign liabilities.

Includes financial derivatives.

Sources: National central bank websites; IMF, World Economic Outlook; and IMF staff estimates.

Foreign assets minus foreign liabilities.

Includes financial derivatives.

A key uncertainty in all these measures is how they would change as a result of joining the euro area. Although membership in a currency union will increase trade links, the implications for cross-country business cycle correlations, at least at a theoretical level, are ambiguous. Krugman (1991c) argues that reduced trade barriers will lead to greater specialization of production as countries utilize their comparative advantage. Artis (1991) suggests that greater production specialization could result as firms with previously scattered production facilities recentralize them once exchange rate risk disappears. Trade would tend, therefore, to be of the less stabilizing interindustry type. In contrast, Frankel and Rose (1996) argue that currency unions increase intraindustry trade and therefore tend to result in more synchronized business cycles. Kenen (2002) suggests that thickening trade channels among members improve the fit of the union’s single monetary policy.

Resolving this debate is ultimately an empirical matter. Empirical evidence for the period since euro adoption is not available, but EC (1999) examines changes in the degree of specialization in production and trade for individual EU countries during 1988–98. Overall, it found no general trend toward increasing specialization, except in Ireland, where specialization in research and skill-intensive industries was strong. Frias, Iglesias, and Neira (1999) obtain similar results for the period 1992–96. These results, of course, are for a pre-EMU period and may change substantially over time.

Adaptability to Asymmetric Shocks

The adaptability to asymmetric shocks after euro adoption will depend primarily on wage and price flexibility and the ability to use fiscal policy counter-cyclically. The latter will be particularly important in response to demand shocks such as credit booms. But for supply-side shocks and many other types of demand shocks, wage and price flexibility will be critical to ensuring adequate competitiveness while avoiding rising unemployment.

Recent estimates suggest that the response of aggregate real wages to unemployment in the CECs is similar to that in mature market economies. Specifically, estimates of the elasticity of pay to unemployment in Hungary and Poland (Kertesi and Köllo, 1997; and Estevao, 2003) are virtually identical to those of Blanchflower and Oswald (1994) in each of 12 industrial countries studied.11 For the Czech Republic, estimates are close to such levels (Munich, Svejnar, and Terrell, 1999). A panel regression for the CECs, using annual data from 1995 through 2002, confirms the similarity of responsiveness of real unit labor costs to unemployment (Box 3.1).

Whatever the responsiveness of real wages to economic conditions, nominal wage flexibility must be adequate to ensure that unemployment does not shoulder the brunt of adverse shocks after euro adoption. Nominal wage flexibility is difficult to pin down in any country but especially in the CECs, where variations in inflation and productivity have been sizable. The aim here, therefore, is to assess it indirectly by examining whether labor market institutions and regulations are likely to produce wage flexibility.

  • Employment protection legislation, which can hamper wage adjustments by increasing job security and making it more difficult for employers to resist wage demands, is a key indirect indicator of the likelihood of wage flexibility. Employment protection legislation in the CECs, except Slovenia, is less strict than in most euro-area countries (Table 3.7).12 Also, labor market reforms since the calculation of these indices have probably improved the absolute and relative positions of Poland and the Slovak Republic.

  • The nature of wage bargaining can influence wage flexibility, with highly centralized or decentralized bargaining tending to produce more flexibility (Calmfors and Driffill, 1988).13 Wage bargaining in the CECs, which occurs primarily at the enterprise level, is less centralized than in the euro area, except in the Slovak Republic, where wage negotiations take place at the national and sectoral levels (EC, 2002c). A related factor is the strength of unions. In all the CECs, including the Slovak Republic, unions are typically characterized as having only a moderate effect on wage bargaining processes (Crowley, 2002).

  • Legally binding minimum wages, which exist in all CECs, may reduce the flexibility of wages, especially if they are high relative to average wages. The CECs compare well with the euro area in this regard. According to Eurostat data for the late 1990s, workers paid the minimum wage in the seven euro-area countries where a minimum wage applies received 40 to 60 percent of the average earnings of workers in manufacturing. In the CECs minimum wages range from 20 to 40 percent of the average wage.

  • Factor mobility can also reduce the need to rely on wage adjustments, but because it is unlikely to be significant in the short run it will not serve as a substitute for wage flexibility (Buiter, 2000).

Table 3.7.CECs and Selected EU Countries: Overall Strictness of Employment Protection Legislation1
Index2Ranking3
CECs
Czech Republic1.711
Hungary1.49
Poland1.610
Slovenia43.2
Western Europe
Austria2.215
Belgium2.113
France3.021
Germany2.518
Ireland0.94
Netherlands2.114
Southern Europe
Greece3.624
Italy3.323
Portugal3.725
Spain3.122
Source: OECD Employment Outlook, (Paris, 1999).

Covering regular and temporary contracts, OECD “version 1, late 1990s.”

Indicators range from 0 to 6, with a higher number indicating stricter employment protection.

Ranking among 26 OECD member countries. Ranking increases with strictness of employment protection.

Calculated by the Slovenian authorities.

Source: OECD Employment Outlook, (Paris, 1999).

Covering regular and temporary contracts, OECD “version 1, late 1990s.”

Indicators range from 0 to 6, with a higher number indicating stricter employment protection.

Ranking among 26 OECD member countries. Ranking increases with strictness of employment protection.

Calculated by the Slovenian authorities.

Box 3.1.The Wage Curve for the CECs

Data from 1995–2002 suggest that the responsiveness of wages to changes in the unemployment rate is about the same in the CECs as in the euro-area countries (see figure).

Following Blanchflower and Oswald (1994), a wage curve equation was estimated for the CECs with annual data from the period 1995–2002:

ln(wt) = a + bt ln ut + dit + et

where ln refers to natural log, wt is the real wage adjusted for productivity (that is, unit labor costs, ULC) in year t, ut is the unemployment rate in year t, dit is a country-specific dummy (for the Czech Republic, Hungary, Poland, and the Slovak Republic, respectively), and et is the error term.

OLS estimates yield the following results:

** significant at 1 percent level; * significant at 5 percent level; Adj.R2 = 0.92; SE = 0.05

At –0.07, bt, the estimated elasticity of real wages with respect to the unemployment rate, is not statistically different from the estimates of about –0.1 for industrial countries found in Blanchflower and Oswald.

Changes in Real Unit Labor Cost and Unemployment Rate, 1995–2002

Sources: Eurostat; IMF, International Financial Statistics; and IMF staff estimates.

Evidence on price flexibility in the CECs is even scarcer. Product market reforms have relaxed market entry and exit conditions, probably contributing to price flexibility (EC, 2002). But it will be crucial to continue to identify structural reforms to promote competition in goods and services markets to bolster price flexibility. This is an area where, as with susceptibility to shocks, endogenous changes after joining the euro area could change the landscape noticeably. Specifically, with increased trade ties and more transparency in international price comparisons, price, and even wage flexibility, could increase.

While wage and price flexibility are essential for adapting to asymmetric shocks with medium- to long-term effects, fiscal stabilizers can contribute significantly to smoothing the effects of shocks with short-term effects. By and large these will operate through automatic changes in activity-related government revenues and expenditures, rather than discretionary fiscal policy actions, which are often hampered by institutional constraints on the design, implementation, and timing of fiscal measures. Even this “automatic” stabilization may occur only with substantial delays if income tax receipts respond to economic developments with a lag.

Rough estimates suggest that automatic fiscal stabilizers in the CECs could be significant, though smaller than in the existing euro area. The size of automatic stabilizers depends on the sensitivity of the budget to economic fluctuations and of output to changes in revenues and expenditure (the fiscal multiplier). Budget sensitivity is largely a function of ratios of revenue to GDP and the nature of the shock (shocks to consumption having a relatively large effect because of the importance of consumption taxes).14 Background work for this paper suggests that a shock to activity equivalent to 1 percent of GDP has an impact on the budget ranging from 0.42 percent of GDP in the Czech Republic to 0.47 percent of GDP in Hungary (see Box 7.1 in Section VII), broadly in line with estimates for the CECs by Corricelli and Ercolani (2003) of around 0.4 percent of GDP. These estimates are somewhat lower than Corricelli and Ercolani’s average estimates for EU countries of around 0.6, consistent with lower revenue-to-GDP ratios in the CECs. Simulations of specific shocks for EU countries with the EC’s macroeconomic model show a budgetary impact of 0.7 percent of GDP for consumption shocks and 0.2–0.3 percent for shocks to investment or exports (Brunila, Buti, and in ‘t Veld, 2002). Estimates of short-term fiscal multipliers, which depend largely on the openness of the economy and vary from zero or less to over one (see Box 7.2), are likely to be similar in most CECs to those in the more open EU countries, and in Poland to the euro-area average. Combining the estimates of budget sensitivity in the CECs with those of fiscal multipliers, automatic stabilizers would be expected to neutralize between 15 percent and 20 percent of the effect of demand shocks on output in the CECs (the higher end of this range reflecting Poland’s less open economy). This is slightly below estimates for the EU that suggest 10 percent of an export or investment shock and 25–30 percent of a consumption shock would be neutralized by automatic stabilizers (Brunila, Buti, and in ‘t Veld, 2002).

Assessing the OCA Criteria Jointly: Cluster Analysis

The broad picture emerging thus far suggests that the CECs fulfill most of the OCA criteria about as well as the noncore euro-area members. But the indicators are not uniform. A cluster analysis provides a more concrete means to assimilate all data. The analysis identifies, according to a well-defined statistical criterion, countries that form relatively homogeneous groups in relation to the joint set of OCA criteria and reveals the key characteristics of these groups (Appendix 3.1).15

Other studies have applied cluster analysis separately to the EU and accession countries, with the following results:

  • On the basis of business and interest rate cycle correlations, openness to trade, employment protection legislation, real exchange rate volatility, and the inflation differentials all vis-à-vis Germany, existing EU members fall into three clusters: a core group (Austria, Belgium, France, and the Netherlands) and two peripheral groups. Relative to the core, a northern peripheral group (Denmark, Sweden, Finland, Ireland, and the United Kingdom) has business cycles that are less synchronized with Germany and labor markets that are less protected; and a southern peripheral group (Italy, Spain, Portugal, and Greece) also has lower business cycle correlation but higher labor market protection (Artis and Zhang, 2002a).

  • On the basis of business cycle correlations with Germany, real exchange rate volatility, openness to trade, and a measure of relative inflation—and covering the eight transition accession countries, plus Bulgaria and Romania—the five CECs fall into one cluster characterized by relatively high business cycle correlation with Germany and trade openness (Boreiko, 2002).

To assess how well the CECs measure up to existing euro-area countries, a joint cluster analysis of the OCA criteria is needed. Clusters are chosen on the basis of variables that capture either susceptibility to asymmetric shocks (correlations of economic activity with Germany, symmetry of structure of the manufacturing sector vis-à-vis the euro-area average, and share of intraindustry trade) or adaptability (labor market flexibility proxied by strictness of employment protection legislation). The correlation of real interest rates with those in Germany is also included to capture the possibility that differing interest rate policies were responsible for similarities or differences in growth rates. The data set and periods covered (generally 1996–2002) are detailed in Appendix 3.1. Clustering was carried out in three steps to examine susceptibility and adaptability separately.

In the first step, countries were clustered according to the subset of OCA criteria that measures susceptibility to shocks only (Table 3.8). The analysis produces four groups.16

Table 3.8.CECs and Euro-Area Countries: Clusters and Underlying Components of Susceptibility Indicators1
Output Correlation2Manufacturing Asymmetry3Intraindustry Trade4
(Mean by cluster)
Core50.7840.071.4
Northern periphery60.8686.254.3
Southern periphery70.6049.836.9
CEC-H and Portugal80.4338.670.4
(Deviation from mean; measured in standard deviations)9
Core50.61–0.330.44
Northern periphery61.002.04–1.09
Southern periphery7–0.310.19–2.64
CEC-H and Portugal8–1.19–0.390.35
Sources: Eurostat; OECD; IMF, International Financial Statistics; and IMF staff calculations.

Sample of 15 countries.

Correlation coefficients of year-on-year monthly rates of growth of industrial output in the country and in Germany, 1996–2002.

Sum of absolute values of the differences between the shares of gross manufacturing output by industry and EMU-weighted average. Data for 2000 (except 1999 for the Czech Republic, Greece, Ireland, Italy, and 1997 for the Slovak Republic).

Manufacturing intraindustry trade as a percentage of total manufacturing trade. Data for 1996–2000, computed for each two-digit (revision 3) SITC product class.

Austria, Belgium, France, the Netherlands, Italy, Spain, and Hungary.

Finland and Ireland.

Greece.

Czech Republic, Slovak Republic, Slovenia, and Poland. Excludes Hungary, which is clustered in the core.

For instance, the average degree of business cycle synchronization for countries in the core is 0.61 standard deviation higher than the average for the entire sample of 15 countries.

Sources: Eurostat; OECD; IMF, International Financial Statistics; and IMF staff calculations.

Sample of 15 countries.

Correlation coefficients of year-on-year monthly rates of growth of industrial output in the country and in Germany, 1996–2002.

Sum of absolute values of the differences between the shares of gross manufacturing output by industry and EMU-weighted average. Data for 2000 (except 1999 for the Czech Republic, Greece, Ireland, Italy, and 1997 for the Slovak Republic).

Manufacturing intraindustry trade as a percentage of total manufacturing trade. Data for 1996–2000, computed for each two-digit (revision 3) SITC product class.

Austria, Belgium, France, the Netherlands, Italy, Spain, and Hungary.

Finland and Ireland.

Greece.

Czech Republic, Slovak Republic, Slovenia, and Poland. Excludes Hungary, which is clustered in the core.

For instance, the average degree of business cycle synchronization for countries in the core is 0.61 standard deviation higher than the average for the entire sample of 15 countries.

  • The core: Austria, Belgium, France, Hungary, Italy, the Netherlands, and Spain. These countries show common features that are far from being fully shared by other countries: (1) high correlation of activity with Germany; (2) relatively low manufacturing asymmetry vis-à-vis the euro-area average; and (3) relatively high intraindustry trade. This core group is similar to that identified elsewhere in the literature.17

  • The Northern periphery: Finland and Ireland. This group distances itself from the core in three main respects: activity is more correlated with that in Germany, but the asymmetry in manufacturing is higher, and the degree of intraindustry trade lower.

  • The Southern periphery: Greece, with two predominant features vis-à-vis the core—lower synchronization of activity and less intraindustry trade. The degree of manufacturing asymmetry is similar to the core.

  • The CEC-H (CECs except Hungary) and Portugal, characterized primarily by a much lower synchronization of activity than the core, but with manufacturing asymmetry and intraindustry trade similar to the core.

The characteristics of the CEC grouping suggest that susceptibility to asymmetric shocks may fall in the future. Specifically, despite weaker activity correlations in CEC-H than in the core, their manufacturing structures are closer to that of the euro area than are those of the northern periphery, and their share of intraindustry trade is well above that in the southern periphery. Observing this pattern in a cluster of four CECs is consistent with the hypothesis that past asymmetric shocks mainly reflected transition shocks, but forward-looking indicators suggest increasing synchronization of activity. That Hungary—arguably farthest along in transition—is clustered with the core lends support to this hypothesis.

In the second step, the set of variables is expanded to capture countries’ capacity for adapting to asymmetric shocks (Table 3.9). The variable added is the degree of labor market protection—a proxy for labor market flexibility. This addition increases the optimal number of clusters. The group of core countries splits into two, with one group (Italy and Spain) showing a higher degree of employment protection than the other (Austria, Belgium, France, Hungary, and the Netherlands). The northern periphery group also splits into two single-country clusters, with both clusters showing low employment protection compared with the core groups. The southern periphery also splits into two single-country clusters (Greece and Portugal) characterized by a high degree of employment protection. Except for Slovenia, countries in the CEC-H cluster show better rankings of employment protection than the groups of core countries. While this property reduces CECs’ homogeneity with the core, it suggests better adaptability to asymmetric shocks and enhances the suitability of these countries for currency union.

Table 3.9.CECs and Euro-Area Countries: Clusters and Underlying Components of Susceptibility and Adaptability Indicators1
Output Correlation2Manufacturing Asymmetry3Intraindustry Trade4Strictness of Employment Protection5
(Mean by cluster)
Core
Austria, Belgium, France, the Netherlands, and Hungary0.7745.072.82.3
Spain and Italy0.8127.568.03.3
Northern periphery
Ireland0.80102.154.61.1
Finland0.9270.353.92.1
Southern periphery
Greece0.6049.836.93.5
Portugal0.3852.261.33.7
CEC-H
Poland, Slovak Republic, and Czech Republic0.4738.672.02.1
Slovenia0.3325.374.73.2
(Deviation from mean; measured in standard deviations)
Core
Austria, Belgium, France, the Netherlands, and Hungary0.55–0.070.56–0.20
Spain and Italy0.76–0.970.130.98
Northern periphery
Ireland0.712.85–1.06–1.71
Finland1.291.22–1.12–0.79
Southern periphery
Greece–0.310.18–2.641.21
Portugal–1.440.30–0.461.51
CEC-H
Poland, Slovak Republic, and Czech Republic–0.94–0.400.49–0.74
Slovenia–1.67–1.080.731.05
Sources: Eurostat; OECD; IMF, International Financial Statistics; and IMF staff calculations.

Sample of 15 countries.

Correlation coefficients of year-on-year monthly rates of growth of industrial output in the country and in Germany, 1996–2002.

Sum of absolute values of the differences between the shares of gross manufacturing output by industry and EMU-weighted average. Data for 2000 (except 1999 for the Czech Republic, Greece, Ireland, Italy, and 1997 for the Slovak Republic).

Manufacturing intraindustry trade as a percentage of total manufacturing trade. Data for 1996–2000, computed for each two-digit (revision 3) SITC product class.

OECD overall strictness of employment protection legislation index (1999). Weighted average of indicators for regular contracts, temporary contracts, and collective dismissals. Each indicator ranges from 0 to 6, and within the sample, from 1.1 (Ireland) to 3.7 (Portugal).

Sources: Eurostat; OECD; IMF, International Financial Statistics; and IMF staff calculations.

Sample of 15 countries.

Correlation coefficients of year-on-year monthly rates of growth of industrial output in the country and in Germany, 1996–2002.

Sum of absolute values of the differences between the shares of gross manufacturing output by industry and EMU-weighted average. Data for 2000 (except 1999 for the Czech Republic, Greece, Ireland, Italy, and 1997 for the Slovak Republic).

Manufacturing intraindustry trade as a percentage of total manufacturing trade. Data for 1996–2000, computed for each two-digit (revision 3) SITC product class.

OECD overall strictness of employment protection legislation index (1999). Weighted average of indicators for regular contracts, temporary contracts, and collective dismissals. Each indicator ranges from 0 to 6, and within the sample, from 1.1 (Ireland) to 3.7 (Portugal).

In the third step, correlations of interest rates with those in Germany are added to help assess correlation of economic activity and the role interest rate policy played in it. The results confirm the preceding cluster characterization, with only a small change in the grouping among core countries (Table 3.10). Not surprisingly, for the CEC-H, the degree of interest rate correlation is low compared with the core countries: coordination of monetary policy with Germany was not the primary objective for these countries as they disinflated in recent years. These results are robust to the choice of variables and the number of clusters.

Table 3.10.CECs and Euro-Area Countries: Clusters and Underlying Components of Susceptibility, Adaptability, and Interest Rate Indicators1
ClustersOutput Correlation2Manufacturing Asymmetry3Intraindustry Trade4Strictness of Employment Protection5Interest Rate Correlation6
(Mean by cluster)
Core
Austria, Belgium, France, Italy, and Spain0.8034.671.82.839.2
The Netherlands and Hungary0.7453.570.52.020.3
Northern periphery
Ireland0.80102.154.61.17.5
Finland0.9270.353.92.157.4
Southern periphery
Greece0.6049.836.93.5-2.5
Portugal0.3852.261.33.7-28.2
CEC-H
Poland, Slovak Republic, and Czech Republic0.4738.672.02.17.0
Slovenia0.3325.374.73.2–8.8
(Deviation from mean; measured in standard deviations)
Core
Austria, Belgium, France, Italy and Spain0.70–0.600.470.500.78
The Netherlands and Hungary0.380.360.36–0.790.06
Northern periphery
Ireland0.712.85–1.06–1.71–0.44
Finland1.291.22–1.12–0.791.48
Southern periphery
Greece–0.310.18–2.641.21–0.82
Portugal–1.440.30–0.461.51–1.81
CEC-H
Poland, Slovak Republic, and Czech Republic–0.94–0.400.49–0.74–0.46
Slovenia–1.67–1.080.731.05–1.07
Sources: Eurostat; OECD; IMF, International Financial Statistics; and IMF staff calculations.

Sample of 15 countries.

Correlation coefficients of year-on-year monthly rates of growth of industrial output in the country and in Germany, 1996–2002.

Sum of absolute values of the differences between the shares of gross manufacturing output by industry and EMU-weighted average. Data for 2000 (except 1999 for the Czech Republic, Greece, Ireland, Italy, and 1997 for the Slovak Republic).

Manufacturing intraindustry trade as a percentage of total manufacturing trade. Data for 1996–2000, computed for each two-digit (revision 3) SITC product class.

OECD overall strictness of employment protection legislation index (1999). Weighted average of indicators for regular contracts, temporary contracts, and collective dismissals. Each indicator ranges from 0 to 6, and within the sample, from 1.1 (Ireland) to 3.7 (Portugal).

Correlation of real short-term money market interest rates, 1996–2002. Detrending by HP filter applied to monthly series.

Sources: Eurostat; OECD; IMF, International Financial Statistics; and IMF staff calculations.

Sample of 15 countries.

Correlation coefficients of year-on-year monthly rates of growth of industrial output in the country and in Germany, 1996–2002.

Sum of absolute values of the differences between the shares of gross manufacturing output by industry and EMU-weighted average. Data for 2000 (except 1999 for the Czech Republic, Greece, Ireland, Italy, and 1997 for the Slovak Republic).

Manufacturing intraindustry trade as a percentage of total manufacturing trade. Data for 1996–2000, computed for each two-digit (revision 3) SITC product class.

OECD overall strictness of employment protection legislation index (1999). Weighted average of indicators for regular contracts, temporary contracts, and collective dismissals. Each indicator ranges from 0 to 6, and within the sample, from 1.1 (Ireland) to 3.7 (Portugal).

Correlation of real short-term money market interest rates, 1996–2002. Detrending by HP filter applied to monthly series.

The cluster analysis indicates that OCA characteristics of the CECs are probably at the strong end of the euro periphery. Hungary is even clustered with the core euro area. The four other CECs belong to a statistically identified grouping that, relative to average noncore groupings, has had low correlations of activity with Germany but strong intraindustry trade and structural similarities to the euro area. Moreover, while past activity correlations may raise concerns, the only available uniform measure of adaptability (employment protection) suggests that all the CECs except Slovenia are likely to have greater nominal wage flexibility than almost any other country in the euro area.

How Useful Is the Exchange Rate as a Shock Absorber?

The OCA analysis starts from the assumption that the exchange rate is a useful tool for addressing asymmetric shocks. The presumption is that because flexible exchange rates can generate rapid adjustment in relative prices, they are good absorbers at least of real shocks, when output losses or overheating would occur in the absence of price adjustment.18 An exchange rate response may not be optimal, however, for other types of shocks, particularly those from temporary monetary or financial sources that do not require a relative price change (Appendix 3.2 illustrates these issues in an IS-LM framework). Thus, the usefulness of independent exchange rate policies declines as the incidence of asymmetric monetary/financial shocks increases. Moreover, the value of flexible exchange rates also depends on how well they respond to real shocks: even when a country is at risk of significant asymmetric real shocks, if the exchange rate is not responsive to them, but rather is buffeted by financial shocks, it would not add to economic stability. In these circumstances, losing monetary policy independence would not be costly. Moreover, in a currency union, asymmetric financial shocks would largely disappear (Buiter, 1995). While conceptually clear, these issues must ultimately be resolved empirically.

Empirical studies of the shock-absorbing role of the exchange rate for industrial countries suggest significant differences between large and small countries. Studies on large industrial countries tend to find that real shocks explain at least the majority of the variance in real exchange rates, suggesting that real exchange rates were reasonable shock absorbers (Clarida and Gali, 1994; Enders and Lee, 1997). Other studies—mostly on small, open economies and distinguishing between the role of the real and the nominal exchange rate—find that monetary/financial shocks explain the bulk of the variability in the nominal exchange rate (Canzoneri and others, 1996; Artis and Ehrmann, 2000).19

Empirical work on the CECs, which is still in its infancy, provides less clear results. Gros and Hobza (2003) find that in the CECs, variability of the real exchange rate has been greater than that of the nominal exchange rate. They argue that this suggests the exchange rate has functioned as a source—rather than a dampener—of shocks because the nominal exchange rate has not moved to offset inflation differentials. They do not, however, analyze the sources of exchange rate movements. Dibooglu and Kutan (2001) find that in Poland nominal shocks contribute significantly to movements in nominal and real exchange rates, while in Hungary the exchange rate response to nominal shocks is more limited. Their sample, however, includes periods with little exchange rate flexibility. Süppel (2003) finds for the Czech Republic, Poland, and the Slovak Republic that when exchange rates are flexible, output relative to that in the euro area Granger-causes the real exchange rate. He concludes that real exchange rates respond to shocks to relative output and help dampen cyclical divergences from the euro area. His study, however, is based on an unrestricted VAR and thus does not identify the structural shocks affecting output and the exchange rate. Kontolemis and Ross (2004) apply a structural VAR (SVAR) with output, prices, and the real exchange rate as endogenous variables to a broad group of transition countries over long periods. They find that the impact of real and nominal shocks on the exchange rate varies across countries.

Further work is needed to directly assess the most common types of shocks and the responsiveness of exchange rates to each type of shock in the CECs. To this end, a VAR was estimated on the real exchange rate, the nominal exchange rate, and output—all defined relative to the euro area so as to identify asymmetric monetary/financial (LM), demand (IS), and supply (AS) shocks. The methodology is described in Box 3.2; details are given in Borghijs and Kuijs (2004) (B and K).20 With an explicit identification of the nature of shocks, it is possible to determine the type of shocks to which exchange rate changes have been linked. Sample periods were chosen to coincide to the maximum extent possible with periods when exchange rates have been flexible so as to avoid distortions from systematic official intervention. For Hungary, where exchange rate flexibility was introduced in mid-2001, estimation had to be carried out over a longer period (mid-1995 onwards); for Poland, where the exchange rate has been progressively freer to move since early 1998, results adding earlier data are also shown; and for Slovenia, where official guidance of the exchange rate has been considerable, no significant change in the regime has occurred since 1993, and estimation begins in 1993.

The impulse response functions (IRFs) produced with the structural VAR are generally in line with the theoretical priors, indicating that the shocks are correctly identified and that exchange rate changes do generate relative price changes.21 Currencies appreciate in response to positive IS shocks, dampening the shock, and depreciate in response to positive LM shocks, amplifying the shock. That the long-run relationships already start to dominate after about 2–2½ years—here as well as in other empirical studies on longer samples—appears to mitigate potential problems due to relatively short sample periods.

To analyze quantitatively whether, on balance, flexible exchange rates have acted as shock absorbers, two questions are asked. The first is whether nominal exchange rates varied in response to asymmetric shocks that affect output. Table 3.11 shows the variance decompositions for the three variables, indicating what proportion of their variability (“forecast error variance”) at a 12-month forecast horizon can be attributed to each shock.22 With the exception of Poland during the short period, the bulk of the variability in output is explained by supply shocks—the share ranging from 62 percent in Hungary to 87 percent in the Slovak Republic. These shocks, however, explain only between 9 percent (the Czech Republic) and 23 percent (the Slovak Republic) of the variability in the nominal exchange rate. Thus, while the variability in output stems heavily from supply-side shocks, these explain relatively little of the variability in the nominal exchange rate. A possible interpretation is that exchange rates have not been as responsive to real shocks as optimally they would be.23 Because this analysis focuses on the volatility of exchange rates rather than trend movements, supply-side phenomena—including B-S effects—that affect exchange rate trends are captured by the constant terms in the equations (which are specified in first differences).

Table 3.11.CECs: Sources of Variance in Exchange Rates and Output

(In percent of total variance)1

Czech RepublicHungaryPolandPolandSlovak RepublicSlovenia
1996/II–2003/II1995/III–2003/II1995/V–2003/II1998/II–2003/III1997/I–2003/II1993/I–2003/II
LMISASLMISASLMISASLMISASLMISASLMISAS
Real exchange rate25932960291235578214731374914454312
Nominal exchange rate3672493353144942928383458192380416
Output433364201862151173185230858731762
Sources: IMF, International Financial Statistics; Eurostat; and national authorities.

The “forecast error variance decomposition” reports the contribution of LM, IS, and AS shocks to the conditional variance of the variables at a forecast horizon of 12 months. Estimated with structural VAR model. For details see Borghijs and Kuijs (2004).

CPI-based, vis-à-vis the euro area.

Vis-à-vis the euro.

Industrial output, relative to the euro area.

Sources: IMF, International Financial Statistics; Eurostat; and national authorities.

The “forecast error variance decomposition” reports the contribution of LM, IS, and AS shocks to the conditional variance of the variables at a forecast horizon of 12 months. Estimated with structural VAR model. For details see Borghijs and Kuijs (2004).

CPI-based, vis-à-vis the euro area.

Vis-à-vis the euro.

Industrial output, relative to the euro area.

Box 3.2.Classification and Identification of Shocks in the Mundell-Flemming (MF) Model

Clarida and Gali (1994) derive a stochastic version of the Obstfeld (1985) open economy macro model with output, prices, and the real exchange rate as endogenous variables. The model exhibits the standard MF results: sticky price and output adjustment, and national outputs that are imperfect substitutes in consumption in the short run, while embodying long-run properties characterizing equilibrium after full price adjustment. Following Blanchard and Quah (1989), in the estimation of the structural VAR, these long-run properties are used as restrictions to identify relative supply (AS), demand (IS), and monetary/financial market (LM) shocks: relative AS shocks are those that have a permanent effect on relative output; and relative IS shocks are those that have a permanent effect on the real exchange rate but not on output; and relative nominal shocks have no permanent effect on relative output or the real exchange rate.1 Conceptually, monetary/financial shocks include changes in relative money supply and liquidity preferences, velocity shifts, varying risk premia, effects of financial liberalization, and speculative currency attacks.

Having identified the shocks by their long-run properties, the short- and medium-run structural dynamics are freely estimated. The contribution of each shock to the variability in each variable can be assessed (with variance decomposition), and impulse response functions IRFs can be generated. If correctly identified, these should show the following:

  • A positive relative supply shock increases relative output. The short-run impact on the real and nominal exchange rates is ambiguous, but eventually prices should fall. In the long run, relative output rises, while the effect on the real exchange rate is ambiguous (Buiter, 1995).

  • A positive relative demand shock increases relative demand. In the short run, the nominal and, due to sticky prices, real exchange rates appreciate, and relative output increases. Eventually, prices increase, and in the long run relative output returns to its old level, while the real exchange rate appreciates if the shock is permanent.

  • A positive relative monetary/financial shock lowers the countries’ interest rate, relative to foreign rates. In the short run, both real and nominal exchange rates depreciate—amplifying the impact of the shock—and relative output increases. In the long run, relative output returns to its old level, and there is no effect on the real exchange rate.

1 IS and LM shocks can be classified together as neutral shocks.

The second question in assessing the absorption role of exchange rates is whether most of the variability in nominal exchange rates is explained by real (IS and AS) shocks or financial (LM) shocks. Table 3.11 shows that the variability in the nominal exchange rate is predominantly explained by LM and IS shocks: discounting the results for Hungary, the contribution of LM shocks is 50 percent or higher in all countries and reaches as much as 80 percent in Slovenia. By contrast, about 40 percent of the variability of the exchange rate in Poland is driven by IS shocks. This suggests that the shock absorption role of the exchange rate has been the strongest in the largest, most closed CEC with perhaps the deepest financial markets. This would also appear to be roughly consistent with the results for industrial countries. In all, these findings suggest that, in the CECs on average during periods with (relatively) flexible exchange rates, (1) the exchange rate has responded little to the shocks that impact the real economy and (2) monetary/financial shocks have contributed significantly to nominal exchange rate variability, particularly in the smaller, more open CECs.

What Do General Equilibrium Models Say?

The two prongs of the analysis so far in this section do not make a compelling case for the need for an independent exchange rate policy in the CECs. Both, however, are based on inferences about the effects of euro adoption drawn from an examination of past developments in each country; and neither allows a direct comparison of macroeconomic stability inside versus outside the currency union. For this, it is necessary to simulate the response of policies and macroeconomic developments to similar underlying exogenous states of the world when a country has adopted the euro and when it has not.

This section examines simulations of stochastic general equilibrium models in order to directly compare outcomes for macroeconomic stability when countries have and have not adopted the euro. These models use a common criterion to judge whether a country is better off inside or outside a monetary union—whether the change in exchange rate regime produces lower variability in inflation and the output gap. The value of this approach is that it can systematically simulate the impact of shocks, calibrated on the incidence of disturbances that have impacted a country historically and on key structural attributes of its economy. This permits a controlled experiment comparing macroeconomic variability inside and outside the currency union. The approach also highlights the contribution of nominal wage and price rigidities to inflation and output volatility in the monetary union when the exchange rate instrument is no longer available.

These models cannot, however, fully endogenize changes in behavior that may attend euro adoption. As such they are subject to the Lucas critique—that behavior of economic agents may be specific to policy regimes. Nevertheless, models such as the GEM, which are built on microfoundations, should be less susceptible to this flaw than systems of reduced-form equations. But still, if adopting the euro promoted greater nominal flexibility, it would probably generate lower actual output and inflation volatility than predicted by the model. These models also focus only on the goal of stabilizing macroeconomic volatility—they do not incorporate benefits of euro adoption that might lead to higher output or rates of output growth through the channels discussed in Section II. These considerations suggest that results from the simulations reported here are likely to understate—potentially by large margins—the net benefits of joining the monetary union.

Results from Previous Simulations

The EC’s study, “One Market, One Money” (EC, 1990), was one of the first stochastic simulations of the impact of EMU on economic performance. Applying the IMF’s MULTIMOD model to a hypothetical monetary union consisting of France, Germany, Italy, and the United Kingdom, it considers how output and inflation variability would differ under alternative intracommunity exchange rate regimes. Differences between regimes are modeled as different interest rate reaction functions (which determine the behavior of exchange rates)24 and different intracommunity exchange rate shocks (which are eliminated in EMU). It considers the impact of a variety of shocks calibrated on historical data to evaluate outcomes for overall macroeconomic stability. The study finds that, compared with a free float, EMU reduces both inflation and output variability. This results from the elimination of asymmetric exchange rate shocks and greater discipline on wage and price behavior in the currency union, offsetting the effects on individual countries of a suboptimal monetary policy.

These conclusions on the strong superiority of EMU over floating rates have been questioned by several authors. Masson and Symansky (1993) argue that the EC’s results hinge on implausibly large risk premium shocks in foreign exchange markets under independent monetary policy.25 Replicating the EC methodology, but using smaller estimates of the risk premium shocks for countries outside EMU, Masson and Symansky find no clear improvement in average inflation variability when comparing EMU with floating exchange rates; they do find that output variability drops under EMU.

In contrast, simulations for the U.K. Treasury’s study of EMU participation find that U.K. inflation and output volatility would be higher if the United Kingdom joined EMU now than if it stayed out (Westaway, 2003). The reduced-form, three-country model, including the United Kingdom, the euro area, and the rest of the world, is calibrated to reflect the contributions of demand, supply, and nominal shocks—symmetric and asymmetric—to output and inflation volatility in the United Kingdom. In the benchmark case, inflation and output volatility are significantly higher (by 36 percent and 20 percent, respectively) inside EMU relative to outside. The size of the effect is sensitive to parameter assumptions. They find that the more active is countercyclical fiscal policy, the lower is the volatility of inflation and output if the United Kingdom were to join the euro area. They also conclude that greater price flexibility in the United Kingdom relative to the euro area would increase inflation variability as prices—rather than the exchange rate—assume the burden of adjustment to asymmetric shocks.

GEM Simulations for the CECs

The effects of joining the euro area on output and inflation volatility are examined in the GEM calibrated on the euro area and an economy with key features of the CECs (Laxton and Pesenti, 2003 and extensions thereof). The GEM has several appealing features for this purpose: it is built on optimizing behavior of firms, households, and the monetary authorities, and it can capture many critical features of the CECs vis-à-vis intraindustry trade with the euro area and productivity shocks differentiated between the traded and nontraded sectors. The model is calibrated (in terms of both behavioral parameters and distributions of shocks) to approximate the Czech Republic as the “typical” CEC and the euro area. Relative to the euro area, the Czech Republic is represented as small (5 percent of euro-area GDP), its nominal wages and prices are significantly more flexible, and its economy is more open to trade. Also, it is more susceptible to shocks, particularly to risk premia and productivity. On this basis, Laxton and Pesenti derive the TEFs—the feasible trade-off between output gap and inflation variability facing the monetary authorities—for the Czech Republic and the euro area when the Czech Republic retains its flexible exchange rate.26 These frontiers are shown as LPH and LPF, respectively, in the top panel of Figure 3.1. That the Czech Republic is subject to larger exchange rate and supply shocks implies that its TEF lies to the right of the euro area’s.

Figure 3.1.Euro Area and the Czech Republic: Feasible Combinations of Output and Inflation Variability Under Alternative Monetary Policy Arrangements1

Sources: Laxton and Pesenti (2003); and IMF staff calculations.

1 Taylor efficiency frontiers and schedules.

This result is compared with a simulation having the Czech Republic as a member of EMU. For the Czech Republic, this is represented as a fixing of the nominal exchange rate, eliminating foreign exchange market risk premium shocks, and a monetary policy based on a Taylor rule defined over the weighted average output gaps and inflation rates of the enlarged euro area. The corresponding TEF for the enlarged euro area (labeled “expanded EMU” in the middle panel of Figure 3.1) is to the left of the original LPF schedule, suggesting that the variability trade-off in the existing euro area will be improved owing to the elimination of risk premium shocks and the presence of some negative covariances between Czech and euro-area inflation and output. But the TEF of the Czech Republic is found to be inferior to that under exchange rate flexibility, implying greater inflation and output volatility in the monetary union (schedule BASEH). Also, the Czech Republic will no longer be able to choose a point on its TEF according to its preferences, but must live with the ECB’s choice.27

While no metric exists for measuring the size of the welfare loss from this shift in the TEF, calculations of a representative increase in volatility resulting from a move from inflation targeting to EMU are possible. Assume that the Czech Republic was initially on its TEF (LPH) at the point that lies on the ray from the origin to point B on BASEH and that after adopting the euro, the ECB’s monetary policy places it at point B. Such a move would lead to a rise in the standard deviation of the output gap from 1.9 percentage points to 2.1 percentage points, and an increase in the standard deviation of inflation around its mean from 1.7 percentage points to 1.8 percentage points. This means that with two-thirds probability, the Czech output gap would lie in the range ±2.1 percentage points around potential under EMU, compared with ±1.9 percentage points under inflation targeting. Inflation under EMU would be in the range 1.2–4.8 percent, compared with 1.3–4.7 percent under inflation targeting.28

Conclusions on the effect of euro adoption on macroeconomic volatility are sensitive in several ways to assumptions on the size of shocks. First, assumptions on the pre-euro-adoption risk premium shocks—which of course are eliminated after euro adoption—are critical. For example, doubling the standard deviation of these shocks produces the schedule LPHALT (middle panel of Figure 3.1), which lies to the right of LPH and intersects BASEH. Inflation and output gap variability in the Czech Republic would then be lower inside EMU than outside as long as the ECB were to choose an inflation-output variability combination in the vicinity of points A or B on the “expanded EMU” TEF. Second, significantly lower idiosyncratic productivity shocks in the Czech Republic (thereby reducing the equilibrating role of nominal exchange rate flexibility in the presence of sticky prices) would also produce a more favorable TEF under monetary union. Third, the results depend on assumptions about nominal rigidities that could be endogenous to euro adoption. Thus, if the Czech Republic were to become as rigid as the euro area after joining EMU, its TEF within EMU would shift right (schedule ALT1H in the bottom panel of Figure 3.1), reflecting the tendency for higher nominal rigidities to raise output variability. On the other hand, more flexibility would tend to shift the post-euro-adoption TEF inward.

The GEM simulations also do not take into account gaps between optimal and actual macroeconomic policies. The GEM assumes that under inflation targeting frameworks, countries achieve their best possible inflation-output gap volatility combinations—that is, they are on their TEF. In practice, if monetary policy is not so successful, feasible variability combinations may be well to the right of LPH, the TEF for the inflation targeting period. Adopting the euro could then shift the actual variability combination left even if the LPH frontier was to the left of BASEH. In other words, moving from a suboptimal combination of output gap and inflation variability into the euro area could improve performance. Moreover, other aspects of euro adoption on the conduct of policies—such as the disciplining effect on fiscal policy and better and more predictable monetary and fiscal policy mixes—could well contribute to reducing volatility.

Ideally, the long-term benefits and costs of euro adoption would be combined in a single metric to determine the net gain or loss. However, no single metric is available, and any conclusions on net benefits must therefore contain a substantial element of judgment on how to weight the possible benefits for trade and growth and the possible costs of greater macroeconomic volatility. Westaway (2003) points out that even large estimates of potential gains from trade may not be realized if euro entry were to result in less stable macroeconomic conditions. For the CECs, however, evidence suggests that the growth dividend coming from an expansion in trade alone will be far larger than for the United Kingdom, while rough calculations of the size of possible increases in volatility seem small. It would therefore seem reasonable to conclude that securing the gains as early as possible, particularly if combined with maximum efforts to improve flexibility of the economy and implement strong fiscal adjustment to support stabilization, would be the growth- and welfare-enhancing course.

Appendix 3.1. Cluster Analysis

The Method

Cluster analysis is a class of statistical methods used for partitioning an observed population sample into homogeneous groups, according to some multidimensional distance function (Everitt, 2001). Since it does not require any assumption on the distribution of the variables in the population, the method is widely used as an exploratory data analysis tool.

The purpose of cluster analysis is to classify the observations in the sample, according to an index of proximity, which is based on some dissimilarity measure calculated for the vector of individual characteristics. During the clustering process, the overall variability of the data set is decomposed into a within-group and a between-group dimension. The objective of cluster analysis is to find the best partition of the sample units—that is, the one that yields the highest value for the ratio of between to within cluster variability.

Since cluster analysis is a descriptive rather than a probabilistic statistical method it cannot be used to test any hypothesis concerning the causal relationship between variables. Different partitions can be obtained as a result of the choice of the variables considered or the number of clusters retained. The number of clusters can be set a priori, and the analysis in this paper starts with a small number of clusters. The number of clusters is then increased until the distance (see below) for each country within each cluster is minimized. This ensures that each cluster is well defined, in the sense that individual clusters have no outliers.

Once the best partition of the sample data has been obtained, a description of the clusters is needed. This can be obtained by analyzing the specific characteristics of each group in terms of the variables used to build the partitions.

The Algorithm

The partitioning in this paper adopts a k-means algorithm based on minimizing the Euclidean distance (that is, the square root of the squared distance) between the points within a cluster relative to those in other clusters. Thus, the algorithm sorts data into “natural” groupings based on similarities within a cluster relative to other clusters. The method allows one to group the sample units in a predefined number of clusters that have the lowest internal variance and the largest intergroup variability. The algorithm consists of comparing the distance of each observation from the average of the predefined clusters. The initial value for the latter is usually chosen randomly among the sample data. At each step, the algorithm assigns an observation to the nearest cluster, and distances are recomputed immediately. The process continues until the results converge to a solution. The iteration stop-rule is based on the size of the reduction in the intragroup variance. When this is sufficiently small, the iteration stops and the final partition is retained.

Formally, let X1, X2, …, Xn be the p-dimensional observations of the n countries to be clustered. The k-means clustering process begins with the specification of the number of clusters initially desired. This number is denoted by k < n. The first k countries are taken as the initial k clusters, each country being its own cluster, so the first observations X1, X2, …, Xk are also the initial mean vectors for the k clusters. The next step is to specify the merging parameter denoted by θc. Then, the distances d(i,j) = (Xi – Xj)′(Xi – Xj) (i, j = 1, 2, …, k, ij) are computed for the k countries. If the minimum distance between any two of these k countries is less than θc, the two countries corresponding to this minimum distance are merged, giving (k–1) clusters for the initial k countries. The mean vector for the merged cluster is computed and also the distances between this cluster and each of the other clusters. If the minimum distance between any of the (k–1) clusters is less than θ, the clusters corresponding to this minimum distance are merged. The process is continued (where each distance computed is the Euclidean distance between mean vectors) until all distances between cluster means are greater than or equal to θc. In this manner, the initial countries are clustered to contain fewer than k clusters. Each of the remaining (nk) countries is placed in the cluster to which its observation has minimum distance from the cluster’s mean vector. Whenever an object is added to a cluster, the new mean of the cluster is computed, and the distance between the means is determined for the new cluster and all other clusters previously formed. Furthermore, the clustering is refined by selecting a second parameter, θr, such that θr ≥ θc, which is called the refining parameter. As each of the additional (nk) countries is considered, and its minimum distance from the current cluster’s means is greater than θr, the country is left by itself as its own cluster, and the process continues. In this manner, the number of clusters may increase, thereby refining the clustering. After all the n countries have been clustered, suppose there are M clusters. The mean vectors of these M clusters are computed, and each object is then reassigned to the cluster whose mean vector it is nearest.

The Data Set

The data set comprises 16 countries29 and five variables, as follows.

  • Correlation of economic activity is measured by the cross-correlations of industrial production with respect to Germany, for the period 1996–2002. As discussed in Boreiko (2002), German industrial production exhibits a high comovement with respect to euro-area GDP, and it is, therefore, a reliable proxy for it. The correlations are based on growth rate of each month relative to the same month of the previous year. Also as discussed in Boreiko (2002), these correlations yield results comparable to correlations calculated for industrial production series detrended using Hodrick-Prescott (HP) filter with the value of the dampening parameter equal to 50,000. Calculations for both correlations are shown in Table 3.12.

  • Manufacturing asymmetry is measured as the sum of absolute values of the difference between the shares of gross manufacturing output by industry and EMU-weighted average. Data are for 2000 (1999 for the Czech Republic, Greece, Ireland, and Italy, and 1997 for the Slovak Republic). The higher the indicator, the greater the deviation from the average manufacturing structure of the euro area. (Source: OECD SNAV database.)

  • Intraindustry trade is measured by Grubel-Lloyd indices, calculated as a share of total manufacturing trade, for 1996–2002. The indices vary between 0 and 1, being equal to unity when trade between two countries is entirely intraindustry. (Source: OECD International Trade Statistics, 2002.)

  • Labor market flexibility is measured by the OECD indicator of strictness of employment protection legislation for the “late 1990s.” It is the weighted average of indicators for regular contracts, temporary contracts, and collective dismissals. For the Slovak Republic, data are not available, and it is assumed the strictness of employment is equal to the level prevailing in the Czech Republic. (Source: OECD Employment Outlook 2002.)

  • Interest rate correlation is measured by the cross-correlation of the cyclical components of the interest rate cycle of a country with that in Germany. The detrending was accomplished by applying the HP filter to monthly series of, alternatively, policy nominal and market real interest rates, with the latter defined as the difference between a short-term market nominal rate and the rate of consumer price inflation. The sampling periods for policy nominal rates are 1993–1998 for the euro-area countries and 1999–2002 for the CECs; and, for market real rates, 1996–2002 for all countries. (Sources: IMF, International Financial Statistics; Eurostat; and IMF staff estimates.)

Table 3.12.Cluster Analysis Variables
Correlation of Industrial ProductionManufacturing Asymmetry3Intraindustry Trade4Employment Protection Index5Interest Rate Correlation6
Growth1Detrended2Market realPolicy nominal
Austria0.890.7738.10.742.30.560.85
Belgium0.630.5036.60.712.50.360.68
France0.860.7743.10.782.80.370.60
Netherlands0.680.4155.10.692.20.370.78
Finland0.920.7370.30.542.10.570.46
Greece0.600.5149.80.373.5–0.02–0.31
Ireland0.800.52102.10.551.10.070.38
Italy0.850.7028.70.653.40.330.37
Spain0.780.6126.20.713.10.330.42
Portugal0.380.1152.20.613.7–0.28–0.01
Czech Republic0.400.3137.00.772.1–0.150.06
Hungary0.790.6251.80.721.70.030.00
Poland0.500.4438.10.632.00.080.31
Slovak Republic0.520.2440.60.762.10.28–0.06
Slovenia0.330.5425.30.753.2–0.090.26
Sources: Eurostat; OECD; and IMF staff calculations.

Correlation coefficients of year-on-year growth rates of monthly industrial production in a country and in Germany, 1996–2002.

Monthly industrial production indices detrended using a HP filter.

Sum of absolute values of the differences between the shares of gross manufacturing output by industry and EMU-weighted average.

Manufacturing intraindustry trade as a percentage of total manufacturing trade. Data for 1996–2000, computed for each two-digit (revision 3) SITC product class.

OECD overall strictness of employment protection legislation index (1999). Weighted average of indicators for regular contracts, temporary contracts, and collective dismissals. Each indicator ranges from 0 to 6, and within the sample, from 1.1 (Ireland) to 3.7 (Portugal).

Correlation of interest rates, 1996–2002. Detrending by HP filter applied to monthly series.

Sources: Eurostat; OECD; and IMF staff calculations.

Correlation coefficients of year-on-year growth rates of monthly industrial production in a country and in Germany, 1996–2002.

Monthly industrial production indices detrended using a HP filter.

Sum of absolute values of the differences between the shares of gross manufacturing output by industry and EMU-weighted average.

Manufacturing intraindustry trade as a percentage of total manufacturing trade. Data for 1996–2000, computed for each two-digit (revision 3) SITC product class.

OECD overall strictness of employment protection legislation index (1999). Weighted average of indicators for regular contracts, temporary contracts, and collective dismissals. Each indicator ranges from 0 to 6, and within the sample, from 1.1 (Ireland) to 3.7 (Portugal).

Correlation of interest rates, 1996–2002. Detrending by HP filter applied to monthly series.

For each variable, the data were normalized by computing for each data point the deviation from the mean for the whole sample and dividing this by the standard deviation. Thus, the values for the mean for each variable within each cluster should be interpreted as the number of standard deviations from the whole sample mean.

Appendix 3.2. Output Response to Shocks Under Different Exchange Rate Regimes—Graphical Analysis

Figure 3.2 depicts a standard IS-LM diagram for an open economy with significant capital mobility. The IS curve (top panel) traces combinations of the interest rate (r) and output (y) that will maintain goods market equilibrium, for given levels of exogenous domestic expenditure (γ¯), the exchange rate (e), and foreign demand (y*). The LM curve traces combinations of r and y that will maintain equilibrium in the money market, for given money supply (M0). The BOP curve traces combinations of r and y that will maintain equilibrium in the foreign exchange market, for given e and y*. The area southeast of the BOP curve depicts a balance of payment deficit.

Figure 3.2.IS and LM Shocks in an IS-LM Diagram

Figure 3.2 (top panel) shows the impact on output of a negative IS shock stemming from, say, a fall in foreign demand, under fixed and flexible exchange rates. Under fixed exchange rates, the IS curve moves inward along the LM curve, decreasing y. The impact of the shock on output is yoyfixed (in the case of sterilization, the LM curve would move inward reducing output even further). Under flexible rates, the IS curve initially makes an inward move of the same size. However, having moved southeast of the BOP curve, the exchange rate depreciates. This moves the IS and the BOP curves out to the new equilibrium with the size of the impact of the shock yoyflexible < yoyfixed. This exercise confirms that flexible exchange rates can play a useful role in absorbing demand shocks.

Figure 3.2 (bottom panel) shows the impact on output of a negative LM shock under the two regimes. Under fixed exchange rates, the LM curve moves inwards along the IS curve, decreasing y. Unless sterilization takes place, the balance of payment surplus will shift the LM curve back to the original position, removing the impact of the shock. Under flexible rates, the LM curve initially makes an inward move of the same size. However, having moved northwest of the BOP curve, the exchange rate appreciates. This moves the IS and the BOP curves inward, further decreasing y. The impact of the shock is yoyflexible > yoyfixed. Under a currency union, there cannot be a (asymmetric) LM shock. Thus, while flexible exchange rates increase the impact of an LM shock on output, compared with a normal fixed exchange rate regime—in line with Buiter (1995), who argues that a flexible exchange rate could be an additional source of shocks, since additional volatility can enter the financial and economic system via the exchange market—a currency union removes asymmetric LM shocks.

Supply and demand shocks are distinguished through the Blanchard and Quah (1989) decomposition whereby shocks are identified as coming from demand if they have no long-run impact on output and as coming from the supply side if they do.

The two-digit SITC classes include several types of intraindustry trade: horizontal trade in similar products with differentiated varieties (e.g., cars of a similar class and price range); trade in vertically differentiated products distinguished by quality and price (e.g., high-quality and low-quality clothing); and trade in one product at different stages of production.

Australia, Austria, Britain, Canada, Germany, Ireland, Italy, Korea, Norway, the Netherlands, Switzerland, and the United States. Despite large institutional differences, the countries studied exhibited great uniformity in wage flexibility.

The Slovak Republic and Slovenia are not included in the OECD study (the latter is not a member).

The original Calmfors and Driffils proposition remains an important reference point, but it has been qualified in at least two ways. First, the hump in the relationship between the degree of centralization of wage bargaining and wage flexibility is flatter the more open is the economy (Soskice, 1990; Calmfors, 2003). Second, the degree of coordination and the breadth of collective bargaining can affect these outcomes as well (U.K. Treasury, 2003b).

Budget sensitivity also depends on the progressivity of the tax system, the relative share of taxation on different forms of economic activity, the generosity of the unemployment benefit system, and the sensitivity of unemployment to fluctuations in output (EC, 2002b).

In this procedure, the overall variability of the dataset is decomposed into within-group and between-group dimensions. The procedure then finds the groupings that yield the highest ratio of between- to within-group variability.

The groups are well defined. By comparing the sample variance of the group means to that within the groups, ANOVA (analysis of variance) tests reject the null hypothesis—that several group means are equal to the population—at the 99 percent confidence level. To ensure that groups are homogeneous, the number of clusters was set with a view to minimizing the distance of each country from the center of its cluster (Appendix 3.1).

The definition of the core varies in the literature, but it virtually always includes Austria, Belgium, France, Germany, and the Netherlands (Taylor, 1995; Bayoumi and Eichengreen, 1996 and 1997; Artis and Zhang, 1998a and -b).

The “New Open-Economy Macroeconomics” theory holds that under certain conditions exchange rate changes do not generate relative price changes because pass-through to import prices is small. Then the exchange rate is of little use as an absorber of asymmetric real shocks (Engel, 2002). However, empirical evidence to date remains supportive of the scope for exchange rates to affect relative prices (Obstfeld, 2001 and 2002).

Other studies compare experiences with fixed and flexible rates across countries (Hoffmaister and Végh, 1995, for Uruguay); or construct a “counterfactual” by comparing the responses to shocks in a standard model and a version with the monetary policy channel “blocked off.” IMF (1997a) did the latter for Finland and found that the shock-absorbing capacity of independent monetary policy was minimal.

The nominal exchange rate is used instead of relative prices (as in the standard Clarida and Gali SVAR). Because the nominal rate will be given up after euro adoption, explicit study of its response to shocks is key. The identification of the shocks remains the same, since the real exchange rate is defined as the product of the nominal exchange rate and relative prices.

See B and K, Figure 3. For Hungary, the IRFs, though different, appear to be consistent with the exchange rate policy of the authorities prior to May 2001.

See B and K (Table A2) for results over the full forecast horizon.

The evidence is not conclusive because as indicated by Canzoneri and others (1996), it would (in theory) be consistent with the exchange rate being such an effective stabilizer that it completely shields output from LM and IS shocks.

In a free float, interest rate reaction functions to domestic output and inflation variability are assumed to have the same form in all countries, and expected changes in the nominal exchange rate are determined by uncovered interest rate parity. EMU is defined as fixed exchange rates with a common interest rate reaction function expressed with respect to average EC output and inflation variability.

They also question whether the target in EMU should be the weighted average of countries’ inflation variability or the variability of average EMU inflation. For the latter, risk premium shocks would increase inflation variability in one country but reduce it in another, implying inflation variability in the aggregate below the sum of individual countries.

The TEF represents minimum combinations of inflation and output gap variability that can be achieved with alternative Taylor rule parameterizations. It is constructed by calculating the Taylor rule outcomes over a continuum of alternative weights in the monetary authority’s loss function defined over variability of the output gap and deviations of inflation from the desired level.

The points labeled A, B, and C on the expanded EMU TEF correspond to output and inflation variability in the Czech Republic shown by points A, B, and C, respectively, on the BASEH schedule.

This assumes that average Czech inflation under both inflation targeting and EMU is 3 percent.

Comprising euro-area countries (excluding Luxembourg—that is, Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, the Netherlands, Portugal, and Spain) and CEC countries (Czech Republic, Hungary, Poland, Slovak Republic, and Slovenia). The results exclude Germany because it serves as the reference point for estimates of correlation of growth rates of economic activity for all other countries.

    Other Resources Citing This Publication