External model validation

General DSGE Modeling
1

Hi everyone,

I wanted to ask you a question regarding comparing my estimated model with the data.

Instead of following the conventional “performance” check, where we compare model with data moments, I was planning to simulate data for variables which were not part of my sample and then plot it against the data counterparts. For example, two of my five observables are investment and consumption. GDP is not part of the sample.

Would it therefore make sense to simulate data for GDP, which is not part of the sample, and then plot it against the output data. My gut feeling tells me that both data series should be very close through construction (i.e. market clearing). Hence, doing this exercise for GDP wouldn’t make sense. Do you agree with this?

Best

Robert

2

If the resource constraint implies that GDP is simply a sum of other observables, then you are not using an overidentifying restriction and the test is not useful. Otherwise, you are truly testing your model.

3

Hi Johannes,

Thanks for your comment. Yes, GDP is simply the sum of both aggregate demand components. Thanks for confirming my initial concerns.

Best

Robert