Calibrating/Estimating DSGE models for two time periods


I was wondering if there is material on estimating DSGE models, whether it is RBC or New Keynesian for the before and after 1980 periods.

There were some papers that discussed results when the model was split up before time and how it fared in compared to data split in the same periods.

I have tried it before, but I wanted to read more into it.


What exactly are you looking for? Regime-switching models that look at different active-passive monetary/fiscal policy combinations (what Eric Leeper does)? Looking at structural breaks in Taylor rules (e.g. Clarida/Gali/Gertler)? Or parameter drifts (e.g. Fernandez-Villaverde et al, Macroeconomics Annual)?

I have made some attempts to simulate the model for the before and after 1980. What I did was have two separate calibrations for before and after 1980. I was wondering if that is the proper way to simulate the model.

Also, according to the manual, there is a way of dating your data. Can this be applied to models using calibration only, or is it strictly for estimation? The number of periods in stoch_simul correspond to the number of periods before the model is in steady state.

If I want to look at the 1948Q1 to 1998Q4, then would I set the number of periods to be equal to 200 periods or I can set the number of periods to be as large as possible.

In the case of adding dummy variables to the model, I looked at the other posts but it does not appear clear to me how to do that. For instance, suppose I want to estimate shock: a = rhoaa(-1) + rhobb(-1) + epsilon and I want rhob = 0 if periods < 100 and rho = 0.5 for periods >100. How can this be done in Dynare.

Last question, if I want to do a Monte Carlo type of simulation or simulate the model 1000 times, can this be done in Dynare, if so, how can I do it?

Sorry, but I don’t understand the exercise you are doing. You need to describe it in a more systematic way.

With stoch_simul, there is no point in dating your data. You just have a bunch of artificial time periods after you start simulations.

For the dummy variables, see [Using dummy variables in dynare). Essentially, you define rhoa and rhob as variables and provide an observed variable series that takes on the values you describe.

If you are talking about stoch_simul, use the simul_replic. However, you did not describe what the uncertainty is about you are trying to simulate.

Hi to all and to Prof. Pfeifer,

I was interested in the estimating ta model with structural breaks in monetary policy. I read some posts.

Can estimation with structural breaks in deep parameters be done in dynare ?
is there documentation for that ?

I read a bit on the link below but and googled further but no info came up.
Thanks for nay guidance yhat might come up!

Hi I very much appreciate if someone can comment on my previous post,

The key question is if I can estimate a model with structural breaks in monetary policy parameters ?


We are currently quite busy with preparations for Dynare 4.5. Therefore, estimation with breaks will have to wait for now as its implementation is rather involved, see

Thanks Prof. Pfeifer!