Automating great-ratio adjustment in repeated quarterly estimation

Estimation
1

Hi All
I would like to estimate my model on a quarterly basis, and would like to update the steady-state ratios (cy iy gy and many others) for each estimation.
I typically do this by saving the numerical values great ratios in the data file and then loading the data file early in the dynare mod file and assigning cy= value1, iy=value2 and so on…, picking the corresponding values from the loaded data file.

I wonder if there is a way to automate this in a block…For bigger, more detailed models, we will have more steady-state ratios to update. Does the preprocessor offer a solution?

Thanks
Reuben

2

I am not sure I understand the question. I usually do this calibration in the steady_state_model-block. See


where I calibrate i_y and k_y

3

Thanks Johannes.
I think you are referring to adjusting the steady state using the relationships in the model.
I am considering the fixing of some of the steady-state ratios from the data (and then letting the other steady state values adjust endogenously). For example, the c/y data average will change slightly as the dataset expands with the addition of new observations.
I would like to automate how we can fix the ratio from the data, for each new estimation,
Currently, I save the data average cynew=0.61 from the data in the estimation datafile along with the other time series. Then in the mod file, I would load the data file, so that the cynew loads in the workspace, and then set cy=cynew, after the parameters are declared.
For the next estimation in the next quarter, I would probably have cynew=0.6 or so, and then i would repeat the same.
For one parameter, this routine is ok. But if 10 steady-state parameters have to be fixed from national accounts data, each time the model is estimated on an updated dataset, this can be a bit painful. So I was wondering if there is a way around, perhaps with the preprocessor…

Thanks
Reuben

4

In that case, the best route is similar. Use a steady state file where you load the data, compute the ratios and assign the parameters.

5

Hi, loading a data file each time the steady state is to be computed is probably a waste of time here. If you need to set a ratio from the data it is better to assign the value in the mod file (before the estimation command). Nothing prevents you from loading the dataset using MATLAB commands in the mod file and programmatically compute the ratio from the data.

Best,
Stéphane.

1 Like
6

When you say

how exactly do you run the “new estimation”? If you rerun the whole mod-file, then what @stepan-a says will work. If you only call the estimation-command, you may have to tweak the approach using the set_param_value-command.

7

Hi Johannes, I will rerun the whole mod file!

8

In that case @stepan-a’s suggestion is the most efficient one.