Working Paper: CEPR ID: DP5633
Authors: Filippo Altissimo; Riccardo Cristadoro; Mario Forni; Marco Lippi; Giovanni Veronese
Abstract: This paper presents ideas and methods underlying the construction of an indicator that tracks euro area GDP growth but, unlike GDP growth, (i) is updated monthly and almost in real time, and (ii) is free from short-run dynamics. Removal of short-run dynamics from a time series to isolate the medium to long-run component can be obtained by a band-pass filter. However, it is well known that band-pass filters, being two-sided, perform very poorly at the end of the sample. New EuroCOIN is an estimator of the medium to long-run component of GDP that only uses contemporaneous values of a large panel of macroeconomic time series, so that no end-of-sample deterioration occurs. Moreover, as our dataset is monthly, New EuroCOIN can be updated each month and with a very short delay. Our method is based on generalized principal components that are designed to use leading variables in the dataset as proxies for future values of GDP growth. As the medium to long-run component of GDP is observable, although with delay, the performance of New EuroCOIN at the end of the sample can be measured.
Keywords: bandpass filter; coincident index; generalized principal components; large dataset; factor models
JEL Codes: C51; E32; O30
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| new eurocoin indicator (ne) (F36) | medium to long-run growth (mlrg) (O41) |
| medium to long-run growth (mlrg) (O41) | new eurocoin indicator (ne) (F36) |
| new eurocoin indicator (ne) (F36) | GDP growth (O49) |
| new eurocoin indicator (ne) (F36) | performance against actual GDP growth (E23) |
| new eurocoin indicator (ne) (F36) | forecasting accuracy (C53) |