Journal Issue

Euro Area Policies: Selected Issues

International Monetary Fund. European Dept.
Published Date:
July 2017
  • ShareShare
Show Summary Details

Real Income Convergence in the Euro Area1

After the European Union was established, the founding members of the euro experienced steady income convergence. However, this convergence process has stalled since the introduction of the euro, except for new euro area members which reduced their income gaps vis-à-vis the founding members until their adoption of the common currency. Convergence of income levels is not a prerequisite for a functioning monetary union, but has been considered an important objective of the European economic integration process. Lagging productivity growth in countries with lower initial GDP per capita is found to be the main explanation for the lack of convergence, suggesting that structural reforms can help to restart the convergence process.

A. Convergence: A Long-Standing Objective of European Economic Integration

1. Convergence of real income levels is not a prerequisite for a functioning monetary union per se, but it is a long-standing objective of European economic integration. Dating back to the Treaty of Rome (1957), real convergence – a gradual and sustained decline in per capita income gaps across euro area countries – has been a major goal of the economic integration process. The Delors Report (European Council, 1989), which together with the 1970 Werner Report sets out the conceptual framework underpinning Economic and Monetary Union (EMU), lists the convergence of living standards as an EMU policy objective, in addition to price stability, balanced growth, high employment and external equilibrium. The Maastricht Treaty, which lays the foundation for EMU, restates this objective, citing its member states’ resolve to achieve the convergence of their economies in its preamble, while Article 2 defines the promotion of “a high degree of convergence of economic performance” as a task of the Union.

2. EMU was expected to foster income convergence through greater trade and capital flows and by creating incentives for reforms at the national level. At the outset of EMU, policymakers assumed that, by eliminating exchange-rate risk and allowing the free movement of goods, services, capital and labor, improved cross-border resource allocation would boost economic growth and help income levels to converge between countries (Aglietta and Brand, 2013). Moreover, the Delors Report predicted that, without recourse to devaluation, the discipline imposed by monetary union would increase incentives for reforms to boost productivity growth. Economic theory supported this thinking. In neoclassical growth theory, the removal of exchange-rate risk and other barriers encourages capital flows to “catching-up” economies with lower capital-output ratios and higher marginal products of capital, thereby boosting investment and economic growth (Blanchard and Giavazzi, 2002; Praet, 2014; Tressel et al., 2014). Likewise, labor could flow from lower-wage countries to higher-wage ones, producing convergence in the marginal product of labor.

3. The founders of EMU recognized, however, that there were forces which could potentially produce divergence. Economic activity could concentrate in more prosperous areas with an agglomeration of human capital and physical infrastructure—a centripetal force also put forward by academic research (De la Dehasa and Krugman, 1992; European Council, 1989). The Delors Report identifies such regional and structural disparities as “grave economic and political risks,” and advocates the “spread of welfare gains throughout the Community” by means of investment programs in areas such as infrastructure, communications, transport and education, to facilitate the equalization of production conditions. EU structural funds were considered an important – though potentially insufficient – instrument in this regard (Emerson et al., 1992; European Council, 1989).

4. The diverging experiences of euro area countries after the crisis have renewed the focus on convergence. The euro area is emerging from a deep crisis that has challenged the ability of its macroeconomic policy framework to deliver stability and prosperity. While countries such as Germany are now well above their pre-crisis GDP levels, for other countries such as Italy, GDP is only expected to return to its pre-crisis level in the mid-2020s. Although support for the euro area remains high, it is highest in countries with high income levels (text chart). Moreover, countries that have experienced high growth since euro introduction are more likely to have seen an increase in support for the euro (text chart). Convergence may indeed be important for the cohesion of the monetary union, as it helps to ensure that the gains from economic integration are shared.

Support for the Euro and Income Level

Sources: Eurobarometer, WEO, and IMF staff calculations.

Support for the Euro and Convergence

1/ Using 2014 GDP data for Ireland to avoid bias from recent data revisions.

Sources: Eurobarometer, WEO, and IMF staff calculations.

B. Real Income Convergence in the Euro Area

5. Convergence analysis considers whether countries with lower per capita income have caught up with richer ones, and whether income dispersion has been reduced. As summarized in Sala-i-Martin (1996), the literature on real convergence distinguishes between β- and σ-convergence. β-convergence occurs when countries with lower GDP per capita grow faster than those with higher GDP per capita, also referred to as catching-up. σ-convergence is observed when the dispersion of countries’ levels of real GDP per capita declines over time, meaning their income levels become more similar. The two kinds of convergence are related: faster growth of countries with lower GDP per capita (β-convergence) is necessary for the dispersion of income levels to narrow (σ-convergence), but not sufficient. Both concepts are therefore considered when gauging the quality of convergence in the euro area.

6. This paper looks at the convergence performance of euro area countries before and after euro introduction. The analysis compares per capita incomes across countries, both for the initial group of twelve countries that adopted the euro before 2002 (the so-called EA-12) as well as the current group of 19 euro area members (EA-19). Luxembourg is excluded from the analysis because its high GDP, small population and large influx of cross-border workers make it an outlier in GDP per capita terms. Per capita GDP at purchasing power parity (PPP) is used to control for cross-country differences in price levels.2 The results confirm those of several recent studies (auf dem Brinke et al. 2015; Barkbu et al., 2016; ECB, 2015, Kaitila, 2014).

7. There was steady income convergence across euro area countries in the decades leading up to the Maastricht Treaty.3 Simple cross-country regressions of average annual per capita GDP growth on the log of per capita GDP show that EA-12 countries with initially lower GDP per capita tended to grow faster than their counterparts with higher initial income levels over the period from 1960 to 1992, implying that there was strong β-convergence (Table 1).4 The dispersion of GDP per capita across countries (as measured by the coefficient of variation) also fell, confirming that higher growth in countries with initially lower incomes produced σ-convergence (text charts).5

8. However, contrary to expectations, income convergence among EA-12 countries slowed after Maastricht and subsequently came to a halt.6 While 23 years is a short time span for convergence analysis, regressions point to a lack of β-convergence of GDP per capita from 1993 to 2015 (Table 1). The time-series plots of cross-country income dispersion (σ-convergence) show slow convergence in the 1990s, a lack of convergence in the first decade of the euro, and divergence since the crisis, reversing the initial narrowing in income dispersion. This recent divergence is found to be statistically significant at the 5 percent level.7

Table 1.β-Convergence Among Euro Area Countries1
EA-12 (excl. Luxembourg)2βR2
EA-19 (excl. Luxembourg)2βR2
memo: EU-28 (excl. Luxembourg)2βR2
Note: *** significant at 1 percent level; ** significant at 5 percent level; * significant at 10 percent level.

Linear cross-country regressions of average annual PPP GDP per capita growth (γ) between time t+1 through T on the logarithm of PPP GDP per capita (y) at time t. Positive values indicate convergence: γi,t+i,t+T = α − β log(yi,t) + εi,t. Regressions were also run for average annual real GDP per capita growth, with essentially unchanged results (i.e. very similar coefficient sizes and degrees of statistical significance).

Luxembourg excluded because it is an outlier with high PPP GDP per capita and a large number of cross-border workers.

No data available for Ireland and the Netherlands.

No data available for Estonia, Latvia, Lithuania, Slovak Republic and Slovenia.

No data available for Lithuania.

No data available for Czech Republic and Lithuania.

Note: *** significant at 1 percent level; ** significant at 5 percent level; * significant at 10 percent level.

Linear cross-country regressions of average annual PPP GDP per capita growth (γ) between time t+1 through T on the logarithm of PPP GDP per capita (y) at time t. Positive values indicate convergence: γi,t+i,t+T = α − β log(yi,t) + εi,t. Regressions were also run for average annual real GDP per capita growth, with essentially unchanged results (i.e. very similar coefficient sizes and degrees of statistical significance).

Luxembourg excluded because it is an outlier with high PPP GDP per capita and a large number of cross-border workers.

No data available for Ireland and the Netherlands.

No data available for Estonia, Latvia, Lithuania, Slovak Republic and Slovenia.

No data available for Lithuania.

No data available for Czech Republic and Lithuania.

σ-Convergence Across EA Countries, 1960-2015

Coefficient of variation, PPP GDP per capita

Sources: WEO database and IMF staff calculations.

1/ Excludes Luxembourg.

2/ Includes Ireland from 1970 and the Netherlands from 1980 onwards.

3/ Includes Lithuania from 1995 onwards.

Statistical Significance of σ-Convergence Among EA-121

Two-sided 95 percent confidence interval around median coefficient of variation

Sources: WEO, and IMF staff calculations.

1/ PPP GDP per capita data. Excludes Luxembourg. Includes Ireland from 1970 and the Netherlands from 1980 onwards.

9. Weak growth in southern euro area countries has held back convergence. Comparing countries’ actual with their expected average annual growth for the period 1993 to 2015,8 it emerges that in three of the four EA-12 countries with the lowest GDP per capita in 1993 – namely Greece, Portugal and Spain – growth fell significantly short of what would have been implied by their income levels and their previous convergence performance. Growth in Ireland, on the other hand, exceeded expectations by far, driving what little convergence there has been among EA-12 countries since the Maastricht Treaty (text chart). Among the countries with higher initial income levels, Italy stands out with a very weak growth performance compared to what is predicted, despite the already modest growth implied by its high initial GDP. While part of the disappointing growth performance is cyclical in nature, it is worth pointing out that even excluding the recent crisis and looking at the period from 1993 to 2007, there is a lack of β-convergence among the EA-12 (Table 1).9

Growth Performance v. Convergence Expectations, 1993-20151

Actual minus adjusted average annual growth in PPP GDP per capita, in percentage points

1/ Convergence expectations are defined as a country’s hypothetical average annual PPP GDP per capita growth implied by its 1993 GDP level and its previous degree of convergence between 1960 and 1992.

Source: WEO, and IMF staff calculations.

10. At the same time, countries that joined the euro area in 2007 or later experienced continued convergence in the run-up to their accession.10 Income differences between ‘old’ and ‘new’ euro area members were large in the 1990s, but narrowed substantially prior to EU and euro area accession of the latter group (text charts). However, convergence for these countries has also slowed since the global financial crisis. Despite this weaker convergence performance in recent years, over the entire period 1993-2015 convergence among countries now in the euro area has been stronger than among EU countries as a whole (Table 1).11

β-Convergence Across EA Countries, 1993-20151

PPP GDP per capita

Sources: WEO database and IMF staff calculations.

1/ As GDP not logged, R2 values shown differ from results in Table 6.

2/ Excludes Luxembourg, and in the case of euro area 19 Lithuania due to missing data.

Convergence of New EA Member States, 1993-2015

Coefficient of variation with EA-12 average1/, PPP GDP per capita. (Year of euro adoption).

Citation: 2017, 236; 10.5089/9781484312353.002.A001

Sources: WEO database and IMF staff calculations.

1/ Excluding Luxembourg.

11. Income disparities persist also at the regional level within euro area countries. Box 1 compares the very different track records on regional convergence within Italy and Germany over the past decade and a half. The two countries feature considerable income variation across regions, but while German regions have converged since 2000, Italian regions have not. Both are countries with extensive fiscal transfers and labor mobility between regions, suggesting that the difference in convergence outcomes cannot be explained by these factors alone. Rather, how fiscal transfers are used as well as differences in policies and underlying economic structures appear to play a decisive role: whereas East German Länder have undergone large-scale infrastructure improvements and structural change spurred by reunification, progress in these areas has been slow in the south of Italy, in part hampered by pervasive corruption (Burda, 2009; Felice, 2013; Iuzzolino et al., 2011). It is interesting to note in this respect that for U.S. states, where data span almost a century, convergence has been strong (Box 2). The convergence has likely been helped by fiscal stabilization transfers and a high degree of labor mobility, though tight economic integration and few obstacles to cross-border activities may also have contributed.

12. Limited productivity catch-up holds the main explanation behind the lack of income convergence between euro area countries. A decomposition of annual GDP per capita growth across euro area countries with high and low labor productivity levels (defined according to real GDP per hour worked in 1999) shows that both groups of countries have experienced a slowdown in total factor productivity (TFP) growth over recent decades. However, the countries with low initial productivity experienced consistently lower TFP growth throughout the sample periods and a more pronounced slowdown. The sharper fall in investment and employment in countries with low initial productivity is also important in explaining the post-crisis divergence in growth trends.

Growth Decomposition

Average annual per capita growth rates in percent, unweighted

Note: Productivity groups defined on the basis of labor productivity. Countries with high initial productivity include Austria, Belgium, Finland, France, Germany, Ireland, and the Netherlands. Countries with low initial productivity include Greece, Italy, Portugal and Spain. No 1990s data available for Austria.

Sources: AMECO, Haver Analytics, and IMF staff calculations.

13. The main adjustment channels – trade, labor and capital flows – did not produce the expected convergence dynamics. Intra-euro area trade is substantial, but has not increased significantly under EMU. Labor mobility continues to be low, with only around two percent of the EU-15 working age population living in another EU-15 country, partly due to language barriers and limited portability of social security claims (Arpaia et al., 2014). Finally, capital flows rose considerably, yet fueled unsustainable credit booms in recipient countries, while foreign direct investment flowed disproportionately to central European, rather than other euro area, countries. The limited catch-up in productivity growth is an underlying factor explaining these developments (Figure 1).

Figure 1.Adjustment Channels in the Euro Area

C. Conclusion and Policy Recommendations

14. Euro area economies’ convergence performance has fallen short of expectations. Following a catching-up process in the decades before the euro, the convergence of EA-12 countries’ per capita incomes has stalled under EMU. New euro members continued to converge up to their accession, though this process slowed during the financial crisis. Convergence among euro area countries is nevertheless stronger than among EU-countries, supporting our conclusion that the euro is not the reason for the slowdown in convergence. Rather, disappointing productivity growth in countries with lower per capita incomes appears to be the key factor holding back convergence.

15. Policy efforts to foster convergence should focus on raising productivity growth. Structural reforms would help improve productivity growth in lagging countries. While the central level can help push productivity-enhancing reforms by deepening the single market and the effective use of EU instruments, the main responsibility for reviving productivity growth rests at the national level. Empirical research has shown that structural reforms play an essential role in boosting productivity growth (Adler et al., 2017). Furthermore, Banerji et al. (2017) shows that labor and product market reforms have a larger impact on productivity growth in countries with low initial productivity levels, thereby providing an important tool to restart convergence.

Box 1.Regional Convergence in Italy and Germany

Significant differences in per capita income levels exist within Italy and Germany. Looking at Eurostat regional data, per capita GDP in purchasing power standards (PPS) varies greatly across German and Italian regions. In both cases, the per capita income level in the richest region is approximately twice that in the poorest. Overall, there is a similar level of regional dispersion, with the coefficient of variation of PPS GDP per capita for Germany’s sixteen Länder just below 0.28 in 2015 and a little above 0.29 for Italy’s five regions (chart).

σ-Convergence within Germany and Italy

Coefficient of variation, PPS GDP per capita.

Sources: Eurostat, and IMF staff calculations.

There has been some convergence across German regions, supported by a rise in per capita incomes in the East German Länder. While data is only available from 2000 and provides an incomplete picture, the regional dispersion of per capita incomes in Germany declined noticeably between 2000 and 2015, indicating σ-convergence. Looking at the coefficient of variation of the East German Länder vis-à-vis the West German average, convergence appears to be driven by the catch-up of the former with their West German peers (chart).

σ-Convergence of East German Länder

Coefficient of variation vis-a-vis West German average, PPS GDP per capita

Sources: Eurostat, and IMF staff calculations.

Italian regions have not converged in recent years. Unlike in Germany, the overall regional dispersion of per capita income in Italy has remained roughly unchanged over the past 16 years. The economically weaker regions in the south have failed to convergence toward the rest of the country, with the result that their coefficients of variation vis-à-vis the central and north Italian regional average are now significantly higher than those between the East and West German Länder (charts).

σ-Convergence of Southern Italian Regions

Coefficient of variation vis-a-vis North and Central Italian average, PPS GDP per capita.

Sources: Eurostat, and IMF staff calculations.

Box 2.Real Income Convergence Across U.S. States

Updating Barro and Sala-i-Martin (1992), U.S. states continue to show clear evidence of convergence. Using data for personal income per capita since 1929, a more adequate time horizon for convergence analysis, we find that poorer U.S. states grew faster than richer ones (β-convergence) and that income dispersion was reduced considerably until the 1970s, remaining flat thereafter (σ-convergence).

β-Convergence Across U.S. States, 1929-2014

(Per capita personal income)

Sources: U.S. Bureau of Economic Analysis.

The U.S. and the euro area are not directly comparable, as the U.S. is a federation, and has greater labor mobility. The strong income convergence across U.S. states may also be supported by fiscal transfers.

Interestingly, however, income dispersion across U.S. states is in the same ball park as dispersion across euro area countries. The coefficient of variation across U.S. states is currently 0.17, above the lowest level of EA-12 countries in 1998 (0.15), but below its current level (0.26). This suggests that income convergence may not be crucial for a well-functioning monetary union, in particular if fiscal transfers are allowed to smooth out the impact of asymmetric shocks.

σ-Convergence Across U.S. States 1929-2014

(Per capita personal income)

Sources: U.S. Bureau of Economic Analysis.


    AdlerG.R.DuvalD.FurceriS.Kiliç ÇelikK.Koloskova and M.Poplawski-Ribeiro (2017) “Gone with the Headwinds: Global ProductivityIMF Staff Discussion Note No. 17/04International Monetary Fund dated April.

    AgliettaM. and T.Brand (2013) Un New Deal pour l’Europe Éditions Odile Jacob.

    ArpaiaA.A.KissB.Palvolgyi and A.Turrini (2014) “Labour mobility and labour market adjustment in the EUEuropean Economy Economic Papers No. 539 dated December. Available at:

    auf dem BrinkeA.H.Enderlein and J.Fritz-Vannahme (2015) What kind of convergence does the euro area need?Gütersloh: Bertelsmann Stiftung and Jacques Delors Institut—Berlin.

    BanerjiA.B.BarkbuJ.JohnT.KindaS.SaksonovsH.Schoelermann and T.Wu (2015) “Building a Better Union: Incentivizing Structural Reforms in the Euro AreaIMF Working Paper No. 15/201International Monetary Fund.

    BarkbuB.B.Eichengreen and A.Mody (2016) “The Euro’s Twin Challenges: Experience and Lessons” in J.A.Caporaso and M.RhodesPolitical and Economic Dynamics of the Eurozone CrisisOxford University Press.

    BarroR. J. and X.Sala-i-Martin (1992) “ConvergenceJournal of Political Economy vol. 100 No. 2 pp. 223251.

    BenczesI. and B.Szent-Ivanyi (2015) “The European Economy in 2014: Fragile Recovery and ConvergenceJournal of Common Market Studies vol. 53 Annual review pp. 162180.

    BlanchardO. and F.Giavazzi (2002) “Current Account Deficits in the Euro Area: The End of the Feldstein-Horioka Puzzle?Brookings Papers on Economic Activity No. 2.

    BurdaM. (2009) “Half-Empty of Half-Full? East Germany Two Decades LaterVoxEU dated November9. Available at:

    De la DehesaG. and P.R.Krugman (1992) “EMU and the Regions” No. 39. Group of Thirty.

    EmersonM.D.GrosA.ItalianerJ.Pisani-Ferry and H.Reichenbach (1992) One Market One Money: An Evaluation of the Potential Benefits and Costs of Forming an Economic and Monetary UnionOxford; New York;Toronto and Melbourne Oxford University Press.

    European Central Bank (2015) “Real convergence in the euro area: evidence, theory and policy implicationsECB Economic Bulletin Issue 5/2015 pp. 3045.

    European Council Committee for the Study of Economic and Monetary Union (1989) “Report on economic and monetary union in the European Community” (Delors Report) Unipub.

    FeliceE. (2013) “Regional Income Inequality in Italy in the Long Run (1871–2001). Patterns and DeterminantsUHE Working papers No. 2013/08.

    ForgoB. and A.Jevčák (2015) “Economic Convergence of Central and Eastern European EU Member States over the Last DecadeEuropean Economy Discussion Paper No. 001 dated July.

    FranksJ.B.BarkbuR.BlavyW.Oman and H.Schoelermann (forthcoming) “Economic Convergence in the Euro Area: Coming Together or Drifting Apart?IMF Working Paper No. 2017/XX.

    IuzzolinoG.G.Pelligrini and G.Viesti (2011) “Convergence among Italian Regions, 1861–2011Quaderni di Storia Economica No. 22 dated October. Available at:–0022/index.html?com.dotmarketing.htmlpage.language=1.

    KaitalaV. (2014) “Transnational Income Convergence and National Income Disparity: Europe, 1960~2012Journal of Economic Integration vol. 29 No. 2 dated June pp. 343371.

    PraetP. (2014) “The Financial Cycle and Real ConvergenceSpeech at the Annual Hyman P. Minsky Conference on the State of the US and World EconomiesWashington D.C. dated 10April. Available at:

    Sala-i-MartinX. (1996) “The Classical Approach to Convergence AnalysisEconomic Journal vol. 106 No. 437 pp. 10191036.

    SørensenB. and O.Yosha (1998) “International risk sharing and European monetary unificationJournal of International Economics No. 45 pp. 211238.

    Treaty of Maastricht on European Union (“Maastricht Treaty”) (1992).

    TresselT. S.WangJ. S.Kang and J.Shambaugh (2014) “Adjustment in Euro Area Deficit Countries: Progress, Challenges, and PoliciesIMF Staff Discussion Note No. 14/7International Monetary Fund dated July.

Prepared by Hanni Schoelermann, based on forthcoming IMF working paper by Franks et al., “Economic Convergence in the Euro Area: Coming Together or Drifting Apart?”

For assessing β-convergence, there is a trade-off between using real GDP data, which is comparable over time, and PPP GDP data, which is comparable across countries. The subsequent analysis uses the latter, to allow the comparison of living standards across countries, but includes robustness checks with real GDP data.

This finding is consistent with other research. See Kaitila (2014), auf dem Brinke et al. (2015), ECB (2015), and Barkbu et al. (2016).

As the regressions are run for a very small sample of 9 countries due to data availability and the omission of outlier Luxembourg, the precise results should be taken as indicative.

The coefficient of variation (i.e. the standard deviation divided by the mean) is typically used in convergence analysis because it relates the standard deviation to the size of the underlying variable across the sample. This allows for meaningful comparisons over time in instances where the underlying variable, such as GDP, displays a clear trend and the standard deviation expressed in absolute units would overstate (understate) dispersion given a rise (fall) in the magnitude of the variable.

This is consistent with auf dem Brinke et al. (2015), ECB (2015) and Barkbu et al (2016), who find that the dispersion of per capita income levels among the initial euro area countries increased in the period from 1999 to 2014.

Bootstrapping is used to estimate a 95-percent confidence interval around a median coefficient of variation.

A country’s expected growth is calculated by multiplying the log of 1993 PPP GDP per capita with the β-coefficient of the EA-12 convergence regression for the period 1960-1992, and deducting the product from the regression constant. This yields the fitted growth rate consistent with the country’s 1993 GDP level under the assumption of steady convergence going forward.

Replicating the expected growth analysis for the period 1993–2007 reveals a mixed picture: all countries apart from Portugal and Italy exceeded growth expectations up to 2007, with Ireland, Finland, Greece and Spain outperforming growth expectations by most.

Of course, this stronger convergence may reflect the convergence demands of the accession process before joining rather than convergence under the monetary union. It may also be the result of selection bias—countries already more predisposed to convergence were the ones who chose to join the euro.

Other Resources Citing This Publication