Business cycle affiliations in the context of European integration

Abstract

We study affiliations for the countries of the European Economic and Monetary Union (EMU) with Germany and the USA, using various business cycle measures derived from quarterly real GDP. These measures are Hodrick-Prescott and Baxter-King filtered series and annual growth rates. By using rolling contemporaneous and maximum (over a short lead/lag interval) correlations, we document increasing correlations of EMU countries with Germany, with these typically being largest during the 1990s. We also document a strong leading role for the USA in relation to these countries in the period since 1993, thereby correcting the fallacy that the European business cycle was disjointed from the USA for most of the 1990s.

Acknowledgements

This research was supported through a European Community Marie Curie Fellowship (programme ‘Improving Human Research Potential and the Socio-Economic Knowledge Base’ IHP-MCFI-99-1), under which the first author was a visitor at the University of Manchester. The second and third authors gratefully acknowledge financial assistance from the Economic and Social Research Council (UK) under grant number L138251030. This research does not reflect necessarily the views of the funding bodies. The authors also gratefully acknowledge the contribution to this research of stimulating discussions with Mike Artis.

Notes

¹ Our criterion for inclusion is that quarterly data are available in the Main Economic Indicators database of the OECD from 1980Q1 or earlier.

² This follows from the simple (and well known) identity 1 − L ⁴ = (1 − L) + (L − L ²) + (L ² − L ³) + (L ³ − L ⁴), where L is the conventional lag operator and the data are quarterly.

³ We use λ = 1600 in the HP filter, which is the conventional value for quarterly data. This is applied after transformation by taking the logarithm.

⁴ Results obtained using quarterly differences are obtainable from the authors on request.

⁵ That is, for each (centred) time period t, we compute the correlation of GDP in a country with German GDP for periods t − 5, … , t, … , t + 5. Among these 11 correlations, the one with the maximum positive value is shown.

⁶ In some instances the BK filter correlations suffer from ‘small sample’ problems, due to the loss of 12 observations at both ends of the total sample when this filter is applied. This is, for example, the case with the correlations of 0.99 found for both Germany and the US with Finland over 1960–1979. Since we have observations for Finland only from 1975, the loss of three years data with the BK filter implies very few available data points in this sub-period. Also, results for the maximum correlations of Germany with the USA in Tables 2A and 2B are not always the same, since the period is measured in relation to dates for Germany in Table 2A and for the USA in Table 2B. In a small number of cases, no maximum correlation or lead/lag is given because all correlations for leads/lags of −5, … , 0, … , 5 are negative.