Bayesian Estimation - Dornbusch Model

[felipe.klein]

Sorry, I am sending it again. I realized I had to make another correction to the data (next post attached). s, the exchange rate, needed to be expressed as a monthly rate of change and now the results make a bit more sense. I would highly appreciate if you could run it and let me know your suggestions, cause I am not 100% sure how I should interpret these forecasts. Thanks very much,

Felipe

[felipe.klein]

Here is the data file "datos" to run the mod. file "thesis". Data is the monthly change of the exchange rate, in logs. I would highly appreciate if anyone could run it and let me know their impressions and how I should interpret the forecast for the next 24 periods.
Thanks very much,

Felipe

[jpfeifer]

Let me clarify my point. Your model variables are in logs and you have done the arbitrary normalization that the steady state of the log nominal exchange rate is 1. Now you have to match this model to the data where the nominal exchange rate is not in logs and its log has not mean 1. The best way is to specify an observation equation that gets rid of the different means (as described in Listing 7 of Pfeifer "A Guide to Specifying Observation Equations for the Estimation of DSGE Models")

That is, you demean the log empirical exchange rate in the Excel file by just subtracting the mean. Then you use an observation equation of the form

Code: Select all

s-steady_state(s)  = s_obs;


telling the model that the observed log exchange rate is one where the mean has been subtracted.

This formulation also should make forecasting easier, because your forecasts are then for the percentage deviation of the exchange rate from its sample mean.

I add a running version.

[felipe.klein]

Johannes, thanks so much, it's very clear now. But, unfortunately, I was not able to run the model.

Matlab goes through the same process again and again showing always the same option...

"Choose one of the following options:

[1] Consider all the endogenous variables.
[2] Consider all the observed endogenous variables.
[3] Stop Dynare and change the mod file.

options [default is 1] = "

[jpfeifer]

Try Dynare 4.4.2. With the Excel file containing more than 1 series it should run now. The error you describe usually happens when people give the data-file the same name as the mod-file.

[felipe.klein]

Great, now it's working! I will analyse the results and let you know if I have further doubts. Again, thanks very much!