I found that there were multiple steady-state solutions of my model. I could fix the initial value at a certain steady state, but I could not know what the endval was. The endval seemed to be very important for the convergence of deterministic simulation, so I wanted to ask how to deal with this problem.
I also want to ask how to detrend the model if there are multiple trend variables.
Of course, terminal values are crucial in foward-looking models. You are looking for a solution given an initial and terminal condition. If there are multiple steady states, it’s even more important to specify them explicitly.
differentiate_forward_vars option can be used for solving this problem?In other words, whether using this option can avoids the question of inaccurate endval?
You need to be more precise with your posts. Is the problem that the values are inaccurate or that there are multiple steady states as originally posted. The latter would be a problem for differentiate_forward_vars, because you have no idea to which steady state the simulation will converge.
I have a problem. My model assumes that there is a trend, which is determined by some endogenous variables. In other words, the trend is not deterministic, unlike the traditional model with a trend, which leads to another steady state when the model is simulated.
In this case, I don’t understand how to remove the trend, so I’m going to use deterministic simulation to solve the problem, but I find that I can’t be sure what EndVAL is.
Once I change some parameters, deterministic simulations fail.
But is the transition path unique?
My model is a macro model of the epidemic, such as Kaplan (2020) and Eichenbaum et al. (2020), which is a macro model nested within an S-I-R-D transmission.The path of the model should be unique, but endVal cannot be determined due to the interaction between the infectious process and the economic variables.
And does your model still feature a trend at the terminal condition? If not, then indeed differentiate_forward_vars should work.
I use differentiate_forward_vars，but the model still not be solved well.Will OCCbin have a great influence on model solution?My model includes ZLB and financing constraints.
Or you can recommend which algorithm to use.
What do you mean with
?
My model is sensitive to parameters, changing some parameters to do counterfactual analysis will not get the solution.
But there are parameterizations where the model solves and the constraints are all honored?
Yes.What bothers me is that deterministic simulation solutions are mostly failures.
This has prevented me from doing research on many issues.
You should try to find out why they are failures. Is it a numerical issues? Does no solution exist in those cases? What drives the model being able to solve in some cases and not being able to solve in others.
Thank you, Professor. I would like to ask if there are any papers about models with endogenous trends.
Please define “endogenous trend”.
The trend term is a function of other economic variables, such as X(t)= f(C(t-1)),X is the trend variable,X is the population growth term,and C is consumptions.
But what makes this a “trend”? It is simply an endogenous object in the model. From what I understand, there will be no trend at the terminal point as there exists a terminal steady state without detrending.