On the fit of the NK model


I am trying to gauge the fit of the NK model by generating data, and comparing it to data sample used in the SW(2007) model.

I have the NK model (using the code available on this website nk.mod), which I run to get observations for inflation, output and interest rate (using the get_simul_replications.m code). Then I run a simple script (co.m) which uploads SW(2007) data, and the data generated by the nk model, and plots them. Then we can compare the two data sets. As you can see I get a terrible fit, which is surprising, as the data set is not even close to each other.

The data file usd contains the SW(2007) data for output,inflation and interest rate (transformed data used for estimation -dy,pin fobs,robs)

Can anyone comment? Appreciate it!
usd.xls (34 KB)
get_simul_replications.m (1.5 KB)
nk.mod (1.78 KB)
co.m (439 Bytes)

For output growth, they are pretty similar. For the interest and inflation rate I am not sure you are comparing the same objects (net vs. gross; demeaned vs. undemeaned)

As always thank you for the reply.

I thought of that, but since I am using the measurement equations(from SW(2007)) and comparing it with the transformed data used in estimation (again from SW(2007)), I thought I’d get a pretty good fit, and won’t have to worry about transformations and all that. Is their something fundamentally wrong with my approach/results?

Using their observation equation to generate artificial observed variables does only work if the model variables used in the observation equation are the same. I haven’t looked at the models to verify this.

Checked all that, still really bad fit…

Where did you get the calibration from?

From smets and waters (2007)

Then you need to check if their calibration is correct for your model. For example, Smets/Wouters have indexing in their model. Does yours as well?

Thanks, I think I figured it out. As always I am grateful for your opinions and comments!