If 1st Order Linearization Does Not Work

Assalamu Alaikum

I tried first order linearization by Uhlig (1999) but Dynare suggest that equation is non-linear.
What to do now?

I shall be thankful if you could answer please.

Can you enter the equations non-linearly into Dynare?

I can try it.
Just let me know that should it be entered with exponential or just in plain manner?
Thank you for quick response.

First try without exp(), only add the model works.

In fact I have worked on the model as:

  1. Basic model with linearization. Its steady state was achieved. But I could not get the posterior.
  2. Added the financial accelrator. Its steady state was achieved. Still I did not try with data.
  3. Trying to add the completefinancial market. Here a single equation is not working. Dynare declared it as non-linear.
    Trying the whole model as non linear would be troublesome at this stage.

What do you mean by “But I could not get the posterior” and then “Still I did not try with data”? Those are two different statements.
It is difficult to say what is the best way for you then. Maybe there is a mistake in that equation then?

With basic model I tried the data. But there was some problem. Dynare produced no results.
Then I kept working on the model and added the financial accelerator. At this stage, results of the stedy state were generated. After that I tried to go ahead with the addition of remaining equations of the financial market and policies. Here one single equation is not working due to the linearity.
My question is that:

  1. should I try to linearize this single equation through second order perturbation?
  2. OR go for the second order perturbation of the whole model?
  3. OR try some other method.
    Your suggested way of starting with non linear is good. I had tried it with exponential. That did not work, then, a month earlier. So I left it there.

I don’t get it. If you linearized manually, but Dynare tells you an equation is not linear, then you did a mistake in the linearization. You would need to fix it.
In terms of approach, if the easy model cannot be estimated, you should debug that before moving on to something more complicated.