I am trying to replicate (without success) the results of the paper Fiscal news and macroeconomic volatility. I am mainly interested in the TFP news.
I have derived the whole model (abstracting from the fiscal side) and used the same observables (apart from the series on taxes). I was expecting to identify a weight of the TFP news in the variance decomposition similar to that in the paper, but it is not the case. Can anyone give me a clue about what I may be doing wrongly?
Thank you very much in advance!
USdataNEWS_SPF_demean.mat (32.0 KB)
LNEWS.mod (14.3 KB)
What is the exact problem you are facing? Running
identification shows various issues.
Thank you very much for your answer prof. Pfeifer.
I do not face a specific problem, I just was expecting to get similar results in terms of TFP news importance since I use the same model and same observables (excluding the fiscal side). When I run identification I get that:
“All parameters are identified in the model”.
Any hint about what I am doing wrong?
Thank you very much in advance.
In Dynare 4.7 I am getting
MOMENTS (ISKREV, 2010):
The rank of J (Jacobian of first two moments) is deficient!
SE_mec1 is not identified!
SE_mec2 is not identified!
SE_mec3 is not identified!
SE_mec4 is not identified!
SE_mei1 is not identified!
SE_mei2 is not identified!
SE_mei3 is not identified!
SE_mei4 is not identified!
SE_mep1 is not identified!
SE_mep2 is not identified!
SE_mep3 is not identified!
SE_mep4 is not identified!
[SE_news1,SE_etx] are PAIRWISE collinear!
Note that estimation results often very heavily depend on the particular model and data at hand. So your finding may not be overly surprising.
I am so sorry, that mod file has switched on some measurement errors for some observables that I removed to post it here. The attached file is the correct one.
You mean that maybe the results are due to the data revisions and similar issues?
LNEWS.mod (14.0 KB)
Your series for
dPI looks strange.