Length of simulations in Dynare++


I am trying to simulate a "Van Binsbergen, J. H., Fernández-Villaverde, J., Koijen, R. S. J., & Rubio-Ramírez, J. F. (2012). The Term Structure of Interest Rates in a DSGE Model with Recursive Preferences.Journal of Monetary Economics, 1–18.’ type production based asset pricing model.

I prefer to do a monthly calibrations, due to requirements of my model. But I am unable to do a large number of simulations (like 100 years or 1200 months or so). It starts generating model series which are either Inf, 0 or extreme numbers close to Inf. I have had success to do for 70 year (840 months) in some cases.

Are there any known issues in dynare++, that don’t allow for large number of simulations, or the problem is specific to my model?

Thanks !

Please use the regular Dynare and enable pruning. Without it, simulations tend to explode.

Hi Johannes,

Thanks for the prompt reply. As you suggested, the .mod file is working now.

However, the theoretical mean of the risk free rate(rf) is coming to be negative after pruning and taking about 5100 monthly simulations. This was not the case when I used lesser simulation in dynare++. The results were close to data.

To double check, and to ensure that this not the issue with my *.mod file, I ran the *.mod file available for "Croce, M. M. (2014). Long-run productivity risk: A new hope for production-based asset pricing ? Journal of Monetary Economics, 66, 13–31. "

Is there any particular reason for this ? If need, I can share the files/results.

Thanks !

This is complicated.
What do you mean with

Either you have the theoretical mean or you simulate it. Also

[quote]This was not the case when I used lesser simulation in dynare++. The results were close to data.

But you reported the simulations in Dynare++ were exploding.

More to the point: With perturbation solutions, there is no way to enforce upper or lower bounds on variables. This is the price you pay for approximation nonlinear functions with polynomials. For this reason, it can happen in simulations that non-negative variables become negative.

Hi Johannes,

Thank you for the reply.
I stand corrected, I meant the simulated value of the mean.

I have been working on my problem lately and issues are much clearer.
Yes, I agree with perturbations, we can’t really ensure sign of the variables. So the results are what they are :slight_smile:.
Also for exploding, I guess the only real solution (if I don’t want to use pruning) is reduce shock size, which has helped.

Thanks again for the help, your reply helped in making my understanding clear.