I will be very thankful if you can help me with following problem:
I want to estimate DSGE model based on existing literature. Now I have to use the real economic data to link the model with reality. If I understood well the key is to set parameters based on real data (author of article said that he did this by hand). I have matlab code, but I think I need to compute analytical version of model.
I’m new in DSGE area and I want to know how I can find parameters based on real data.
Any ideas/recommendations will be very appreciated.
I don’t know if I get the point of your question. Do you already a have the model calibrated, or is this what you want to do?
Calibration means that we tell the model all the parameter-values and use it for instance to simulate different shocks and their propagation through the economy and time. To do this we normally use economic intuition (like the different shares of GDP) as well as, like you mentioned, existing literature. If you have the model from some paper, why not just start with its parameters?
Estimation is that we take the model to the data and let the data show us what the most likely values for a subset of the parameters we are interested in are. But to do that a working calibrated model is highly recommended.
From your question I seem to believe that you want to calibrate your model. What code do you have, and from where is it? Is it ‘pure’ Matlab code or for Dynare, ie are the files only .m or .mod?
For the compute analytical version of the model, do you mean the steady state?
If you downloaded Dynare, there are some very good examples inclueded of how it works and what you can do with it.
First of all, thank you very much for the prompt response.
I want to calibrate and simulate the model for my own data, the model is for other economy and I can’t use those parameters.
I have Matlab code (files are .m) from author of the article and I understood from him that the code provided doesn’t link the real data with model. I have to compute the steady state by hand and set the parameters to match the real data.
I attached the code, it’s from the paper’s author. matlab_model (1).zip (21.4 KB)
Do you want to use Dynare to work on that model? Because the author of the paper did not use it. Files for Dynare end with a .mod and have a specific way of how variables, equations, etc. have to be entered.
So I guess you first have to decide how you want to work on that model, if you decide on Dynare I suggest you look up some introductions online on how to handle it.
For the calibration part, like I mentioned above, you typically first use values from the literature. This is also what the author in your files did, as you can see in the CALIB_NCC.m line 25 or 32. There are references to, I guess, other papers in that literature.
So if you use the exact same model just for different country data, or something like that, I would suggest you look at the literature for that country where similar models have been used and check what parameter-values were used there. Later when you do estimation, ie bring the model to the data, you also want to have some reference-points to compare your outcomes to.
Thank you very much, it was more than clear regarding model parameters.
In this matlab code model I didn’t find any reference to model data, and ,bring the model to the data" is very difficult step for me now. That was the reason why I’m asked the author of the article how to link the model to the data.
I don’t know how and whether the author used the code for estimation.
As said, if you want to use Dynare you should read some instructions. With this software you can then do estimation, be it Bayesian or ML, of your parameters of choice.