Wednesday, February 23, 2011
A former student from my time series course sent me the following graph:
He has used H-P filter and the Christiano-Fitzgerald approximation to the band pass filter to get the cyclical component of real GDP from 1987 up to present day. He is wondering if he could say, by looking at the this graph, that the output gap has closed.
I prefer to look at GDP together with its trend, but from the graph, it does appear that the output gap has closed. In fact, you get a similar picture if you decompose trend component of GDP using just the H-P filter:
The problem is that this is a mechanical and unrefined methodology, which can lead to wrong conclusions. Cevdet Akcay and Eren Ocakverdi have written a technical note to explain why this is the case. If you just want to get the intuition, have a look at this post...
BTW, because of these limitations, my argument that the output gap has closed down is not merely based on doing this exercise on Eviews. I use as much as art as science, such as monitoring of demand conditions:
At the end of the day, even if you spend a month with time series, it is going to boil down to your subjective opinion...