Dear Prof. Pfeifer,
As Master’s Project, I am working on a modification to Gerali et al. (2010). Specifically I change the utility function to the usual CRRA, and make impatient agent constrained from the formal banking sector by assumption, and instead they borrow informally from Patient household (I am trying to capture whether moneylending and informal banking activity, more often seen in developing countries all over the world) has an impact, by doing counterfactual analysis (one with moneylending, one without) but this model cannot find the steady state.
In order to get a higher interest rate on the moneylending activity, I added endogenous default on these informal loans.
I get the following error:
Impossible to find the steady state. Either the model doesn’t have a
steady state, there are an infinity of steady states, or the guess values
are too far from the solution.
I have been looking for solutions through the forum but given the size of the model, and my impending deadline, I now took the route of writing you directly because I’m reaching the desperation point.
Please find attached my mod file. Any suggestion will be well regarded. Also I would like to disclose the fact that I am a beginner using Dynare but I am a completely risk neutral agent, hence the challenging model selection.
Thesis_V2.mod (28.6 KB)
I am sorry, but this is one of the models that requires a lot of work to modify it. The reason is that no analytical steady state is provided, but just initial values. That makes the mod-file extremely non-robust. I can only recommend going for an analytical steady state. That being said, I would for now focus on
resid: The initial values for the steady state of the following variables are complex:
You are taking the log of a negative number here.
I got rid of all the mk’s and leave the epsilons for simplicity in indexation. However, when you point at an analytical steady state, what do you suggest? Because being such a large model, I was kind of avoiding having to manually compute all SS’s. Is there any way to loop over initial values by linking the modfile to a matlab file with a vector for initval for each variable?
Also, thanks for your prompt reply. I noticed how hard to modify this mod file was waay after being quite advanced with the thesis.
Hello Prof. Pfeifer,
I have simplified the model by eliminating the endogenous default and I can’t still find the steady state. Any suggestions? I have tried everything, right now running 10,000,000 iterations.
According to my experience the issue is often about debugging still hidden problems. If there is only a slight mistake somewhere, a steady state may not exist. In this case, you can try infinitely many starting values and will not be successful.
Unfortunately, there is no way to distinguish this case from one where the problem truly is just with starting values. That is why I recommend using analytical steady states. It will point you immediately to problematic equations.