The RATS Software Forum

Respected Sir

I am trying to use Factor Augmented Vector error correction approach but i am unable to find codes for it. Is it equivalent to finding factors and then using the johansen's method of cointegration on those factors. please help as it is not very clear. A description and application of this method is given in Banerjee, Anindya & Marcellino, Massimiliano, 2008. "Factor-augmented Error Correction Models," CEPR Discussion Papers 6707, C.E.P.R. Discussion Papers.

Regards
Deepika

I agree that it's not as clear as it could be about how they actually do the analysis. The key problem with this, though, is that they rely on Bai(2004) for choosing the number of factors, and it comes up with the rather absurd conclusion that a set of 106 series with widely differing levels of trend and smoothness are best explained by a single factor. To me, that sounds like Bai's IPC criteria aren't tuned properly for practical use. Rather than address that, they simply say---Bai's criterion picks one, we'll use eight.