Anticipated and unanticipated permanent shock with linear_approximation

Dear all,
I have a question on how shock is implemented in perfect foresight simulation.

I want to implement a permanent tax in my model. I know that there are many topics about this. However, due to the complexity of my model and I don’t need to preserve the nonlinearity of the model, I find out that there is an option stack_solve_algo=0, linear_approximation.

However, I am not sure if I use this, my shock is an unanticipated or anticipated shock. Secondly, I want to simulate a ‘transition dynamic’. I am not sure if this is the correct way to do it. Thirdly, will it give me a different results compared to more common way with initial, endval(with the value of tax) and perfect_foresight_solver ?

Thank you so much for all kinds of help.

The linear_approximation-option will preserve the characteristics of perfect foresight simulation, but simply use a linear approximation of the model instead of the full nonlinear one. All shocks will be perfectly anticipated as it’s still perfect foresight.

Thank you so much for your quick reply.
So if I understand correctly, whenever we use perfect foresight, the shock is anticipated.

Let’s say, I implement the shock from period 5:100. But this means the shock is anticipated from period 1. Do I understand correctly?

Is there any way to have permanent unanticipated shock?

  1. Yes, you understand that correctly.
  2. In the unstable version (Dynare 6), there is a perfect_foresight_with_expectation_errors_*-command to allow for that. See
1 Like

Many thanks, this surely helps a lot.