Revisiting the debate on the Eurozone crisis: causes, clustering periphery and core, and the role of interest rate convergence

ABSTRACT

Explaining the Eurozone crisis requires explaining the origins of the external imbalances until then. The paper divides the literature arguments into three ”fundamental causes”, not mutually exclusive: a competitiveness problem, North-South flows, and excess of public and/or private spending. Within each of these causes, we find divergences between authors in the literature and identify a set of variables correlated with the accumulation of external imbalances before the crisis. These variables help to create clusters of Eurozone countries and separate core countries from the periphery. We then develop an original argument, that the convergence of nominal long-term interest rates in the periphery cluster countries, relative to the core cluster countries, between 1996 and 2007, was the trigger for the three ”fundamental causes” mentioned. We find Granger-causality between this convergence and subsequent (after four quarters) annual variations in the quarterly current account balance.

KEYWORDS:

• Eurozone crisis
• external imbalances
• convergence
• cluster analysis
• granger causality

JEL CLASSIFICATION:

• F32
• F41
• G01

Disclosure statement

This article was developed in the context of the author's PhD in Political Economy, interdisciplinary doctorate, which results from a partnership between the following university institutions: ISCTE-IUL, ISEG-UL, and FEUC. Part of the PhD tuition fees were financed by the author's employer, Banco de Portugal. I am grateful to Arslan Razmi, Francisco Louçã, Michael Ash, and Sofia Vale for their helpful comments and suggestions. All remaining errors are the author’s sole responsibility. The analyses, opinions and findings in this paper are not necessarily those of the author´s employer.

Notes

1. The empirical literature excludes the connection between public debt and sovereign debt spreads since spreads were changing in the same direction before the crisis, whatever the public debt ratios.

2. Commonly defined as the ratio between the nominal wage per worker and the labour productivity or, in the same way, the ratio between nominal wages and total output (GDP).

3. The same authors computed that reducing ULC in Germany, Greece, Italy, Portugal, and Spain has little effect. Input costs and the mark-up would be the other variables strongly affecting the final price.

4. According to Chen, Milesi-Ferretti, and Tressel (2013), rapid income growth in commodity-exporting countries benefited countries like Germany, exporting commodities in high demand by oil producers.

5. TARGET is an acronym for Trans-European Automated Real-time Gross Settlement Express Transfer System.

6. Usually, the financial counterpart is recorded both in the ”Currency and Deposits” instrument and ”Currency in circulation” (Central Bank´s balance sheet), as an increase in net acquisition of financial assets.

7. Regression line: cab = 0.055 − 0.049*ltir. The slope coefficient is statistically significant (p-value: 0.001). Data retrieved from the European Commission (AMECO) and ECB.

8. According to IMF, is a measure of the value of a currency against a weighted average of several foreign currencies, divided by an index of unit labour costs. Regression line: cab = 9.412 − 1.759*rerulc. The slope coefficient is statistically significant (p-value: 0.04). Data retrieved from IMF. See Figure A1 in the Appendix section.

9. According to the KLEIM database and looking only at the countries at the tail end of the chart, these are the sectors with the largest positive variations, in terms of share in employment: construction and finance, insurance, real state, and business services (Ireland); real state, renting and business services, hotels and restaurants, community social and personal services (Portugal and Greece). Regression line: cab = 58.52 + 19.23*manu. The slope coefficient is statistically significant (p-value: 0.028). See Figure A2 in the Appendix section.

10. Regression line: cab = −25.931 + 8.874*china. The slope coefficient is statistically significant (p-value: 0.028). Data retrieved from UN Comtrade Database. See Figure A3 in the Appendix section. We also tested other explanatory variables related to external competitiveness, with no statistical significance: ”Changes in Real Effective Exchange Rate based on CPI” (data from IMF); ”Changes in real labour productivity per person/per hour worked” (data from IMF); ”Changes in unit labour costs based on hours worked/on persons” (data from IMF); ”Changes in the Economic Complexity levels” (data from The Observatory of Economic Complexity).

11. Regression line: cab = 13.203 − 2.416*travel. The slope coefficient is statistically significant (p-value: 0.005). Data retrieved from Eurostat database. See Figure A4 in the Appendix section. We also tested other explanatory variables related to the financial account, with no statistical significance: ”Changes in debt securities by country of issuance, in domestic currency in international markets” (data from BIS); ”Changes in stocks of debt securities issued (liabilities) cross border by all sectors” (data from BIS); ”Changes in financial institutions loans (net assets), as a percentage of GDP” (data from Eurostat).

12. Regression line: cab = 4.303 − 26.919*gov. The slope coefficient is statistically significant (p-value: 0.019). Data retrieved from Eurostat. See Figure A5 in the Appendix section.

13. Regression line: cab = 22.183 − 0.695*credit. The slope coefficient is statistically significant (p-value: 0.038). Data retrieved from BIS website. See Figure A6 in the Appendix section. We also tested other explanatory variables related to excess of investment over savings by public/private sectors, with no statistical significance: ”Changes in cumulative public debt, as a percentage of GDP”; ”Changes in Households (S14+S15) net assets”; ”Changes in non-Financial inst. (S11) net assets; ”Changes in Financial inst. (S12) net assets”; ”Changes in Government (S13) net assets” (all data from Eurostat).

14. Regression line: CA = 1.94*(TIME) − 2.53*(TREATED) − 5.97*(DID), R^2 = 52.7%. In which TIME is a dummy variable that takes the value equal to one if it is greater than 1996; TREATED is a dummy variable that takes the value equal to one if the country belongs to the periphery cluster; DID is a dummy that takes the value 1 if it is after 1996 and the country is peripheral, resulting from the product of the last two variables. All slope coefficients are statistically significant at 1% level. Luxembourg was excluded from this first exercise due to lack of data for this time series.

15. The null hypothesis of the Granger-causality test is that the lagged value(s) of the independent variable do(es) not Granger cause the dependent variable. The test statistic is F-distributed for fixed effects. For simple VAR estimation, the test statistic is χ-distributed. Standard errors are clustered along the cross-section for fixed effects estimations, and a constant term and time dummies were included.

Additional information

Funding

The work was supported by the FCT – Fundação para a Ciência e Tecnologia (Portugal) [2021.05858.BD].