My several variables in the model is exogenous, the change of them follow their own deterministic nonlinear change process, but I want to keep the process equation of nonlinear properties, so I choose to declare them varexo_det, but in the model block to write their process equation, dynare reported error: equation number is greater than the number of endogenous variable.But I don’t want to linearize the process equation by writing these exogenous variables into the var declaration. What should I do?

Sorry, but I don’t understand. How can you have more equations than variables? An equation for a process means that the result of the equation is endogenous, not exogenous. In particular, agents know the structure of the process in RE modeling. That is inconsistent with `varexo_det`

, where the actual realizations are perfectly known.

Thank you, Professor, I understand what you mean, because I also put the process equation of exogenous variables into the model module, but I did not declare these variables in var, so the number of equations is too much.After listening to what you said, I understand that the nonlinear nature of these processes should not be retained. Thank you for your advice.

I want to ask you a question, what is the main variable in varexo_det, so let’s say I have a variable a that satisfies a is equal to b*a(-1), b is a constant, and this belongs to the actual realizations are perfectly known？

If `b`

is a constant, then it’s a parameter. Only `a`

would be a variable.

I want to ask another question. In the case of stochastic simulation, the steady state of my model is not unique. In other words, there is a permanent shock to the economy to another steady state. Is there a way to solve it?Or can we use MATLAB code to solve this model by specifying that the steady-state deviation variable of the first order linearization corresponds only to the initial steady-state.

See e.g. Unit root, prespecified S.s., large deviation less accuracy?

and Model simulation with unit root

In a nutshell: unit roots pose not problems for the solution around the initial steady state, but may be problematic if your simulation wanders far away from that steady state.

Does dynare’s use of stochastic simulation necessarily require that all endogenous variables in the model return to the initial steady state?

Generally not. But the precise answer depends on what exactly you are trying to do.

Thank you, professor. I’d like to ask you another question. What causes the stationary solution in general?I used to think that we couldn’t find the stationary solution when the model was subjected to a permanent shock that caused multiple stationary states

Non-stationary shocks will generally cause non-stationary model responses. However, one can ofter render the model stationary by appropriately detrending them. See e.g. https://github.com/JohannesPfeifer/DSGE_mod/blob/master/Aguiar_Gopinath_2007/Aguiar_Gopinath_2007.mod

Thank you professor, I will study how to detrend the model carefully.