Article ID: | iaor20043635 |
Country: | Netherlands |
Volume: | 19 |
Issue: | 4 |
Start Page Number: | 685 |
End Page Number: | 700 |
Publication Date: | Oct 2003 |
Journal: | International Journal of Forecasting |
Authors: | Holden Ken, Duarte Agustin |
Keywords: | time series & forecasting methods |
In this paper the Hodrick–Prescott filter is used to decompose real GDP for the G7 countries into cyclical and trend components. The resulting series of cyclical components are then examined for static relationships, using correlations and graphs; long-run relationships using autoregressive-distributed lag models; and short-run relationships, using error-correction models. The main result is that the patterns of cyclical behaviour changed following the oil price shocks in the 1970s. Since 1980, cyclical fluctuations have been smaller as a result of a decline in synchronisation of the cycles in the G7. Two separate cycles seem to be developing since 1990. One is for Germany, Italy and France, whilst the other is for the US, UK and Canada. Within each of these groups there are both long-run and short-run relationships between the cyclical components of GDP.