I have a DSGE model that incorporates three variables that are already forecasted (from year 2020 to 2040) from a bottom-up engineer model in Gams.

I would like to know if it is possible to fix the value of these “exogenous” variables(from the forecasted data extracted from Gams) for all the years (2020 to 2040) I am estimating the endogenous variables of my DSGE model; so these variables would be “fixed” variables for Dynare.

I am using a standard RE DSGE model with forward-looking agents. The values are known today indeed.

More specifically, I have 2 scenarios of evolution of the electrical mix simulated by an engineer model. These 2 scenarios, obtained from Gams, give me the total electricity supply that can be consumed by households (-> electricity consumption in the utility function and and budget constraint) and costs of the electricity system implied by that supply (-> electricity investments in the resource constraint). These scenarios are known from agents. These two scenarios are incorporated in a stochastic GE. Hence, the electricity investments and electricity supply are kind of “fixed” and “deterministic” for the period of simulation of my model.

I also have a similar question which I was going to write today. I am glad that someone raised it. In the usual Bayesian estimation routine, the number of observable should be at most equal the number of shocks. I have six shocks. Out of these six shocks, I have estimated two shocks already from an external model. I don’t need dynare to estimate these two. I would like to input these two shock series just as data to dynare. Thus for dynare I only need to give four observable for four remaining shocks, The remaining two shocks should come from an external data file. I am struggling to do this. Any help?

1- Maybe the command initval_file could do the job, as it specifies a path in a deterministic setup for all endogenous and exogenous variables ?

2- Or the conditional_forecast command and conditional_forecast_paths command since it uses a given constrained path of the future endogenous variables ?

I’m reading that “When the model also contains deterministic exogenous shocks, the simulation is computed conditionally to the agents knowing the future values of the deterministic exogenous variables”. Can stochastic shocks be incorporated as well?

Sorry, but this is a more complicate issue, so I need more time to think about it. Maybe @stepan-a has an immediate idea. My current hunch is that you cannot easily do this in Dynare.

I think it would be enriching for Dynare to have such a feature: being able to integrate “out-of-the model” forecasted variables, since more and more economists are trying to combine scientific models and macroeconomic GE models.