Strange collinearity error


#1

Hi,

I have wrriten a log linearized model in dynare however i cannot estimate the model.model_diagnostics show that equation 16 and 17 are colinear but it seemes somehow strange since equation 16 is

pint+x(-1)=pint(-1)+x+piintf;

and equation 17 is AR(1) process for foregn country’s inflation piintf as follows

piintf=rropiintf*piintf(-1)+veppiintf;

how can I fix it?

question.mod (4.0 KB)


#2

The price level in New Keynesian models under inflation targeting is not unique/stationary. For that reason, there will be a unit root and hence collinearity. That is normal. But you are missing 2 unstable roots. That suggests that there are timing errors in your model.


#3

Dear Johannes;

Many thanks for your insightful response, but I should say that in this model pint is not a nominal price level, it is real price of some kind of imported price quoted in importer country’s currency deflated by consumer price index. In this model Law of One Price holds so pint changes due to 1- foreign inflation(piintf in code) 2- domestic inflation(pi in code) 3- growth of exchange rate(dh in code). x is an helping variable i.e. exchange rate over CPI so its difference covers item 2 and 3 (2- domestic inflation 3- growth of exchange rate), which is exactly can be seen in equation #19.

Would you mind helping know whether Dynare have problem with this kind of modeling and what is the correct way of modeling.

Finally, about missing unstable roots, it is mostly due to specific conditions that I wanted to check about equation #39. Rank condition can be satisfied by changing that equation in the following form.
gm=rrom*gm(-1)-pi-y+vepm;

Again many thanks for your comments.

Are there any new suggestions for me?


#4

Is it intended that your model’s real variables have a unit root? That is very unusual.