I have a question regarding nocheck command. I am trying to replicate a textbook model. The authors provided replication codes on their website. They used Matlab and their own toolbox. I tried to replicate the model in dynare using their exact specification and calibration. The model only works under nocheck command. Without nocheck command, it gives the following error:

Error using print_info (line 75)
The steadystate file did not compute the steady state
Error in steady (line 104)
print_info(info,options_.noprint, options_);
Error in mxnsgu (line 487)
steady;
Error in dynare (line 278)
evalin(‘base’,fname) ;

When I check the residuals of the static equations, it has nonzero residuals for 4 equations. With nocheck command, there is still nonzero residuals but dynare replicates the moments and impulse responses in the book.

I searched in the manual and forum about this command and as far as I understand, this command blocks dynare to compute the steady-state and use the values provided in the steady_state_model block. Am I correct? What is the use of nocheck command exactly? And is it safe to make modifications on the model with nocheck command?

The nocheck option is generally used in models where you have random walks with drift, ie equations of the form:

a_t = \alpha+ a_{t-1}+\varepsilon_t

There is no steady state for a, except if \alpha=0 in which case there is an infinity of deterministic steady states. In this situation (where the steady state does not exist) you can provide an arbitrary value for a in the block defining the steadystate or in a steady state file routine. This is safe as long as the elasticities and the ratios appearing in the linearized form of the model do not depend on a. Since there is no steady state, any value you provide for a is wrong, hence the nocheck option.

Coming back to the original question, it is rarely ever safe to use the nocheck option unless you explicitly know that you need it. In your MXN model, a steady state exists.

@jpfeifer@stepan-a@MaximeBouter thank you all for your responses. I have been working on the MXN model for 5 months and I am stuck… What would you recommend me to check? I believe the problem is not a drift? Is it the calibration? Or the model equations? Since the model has a steady state in Matlab and Uribe toolbox, I must have made a mistake somewhere… Any idea would be really appreciated.

Here is my earlier question where I explained the model and my problem.