The RATS Software Forum

I am trying to estimate a NK DSGE model. I have a question on the difference between "steady states" and "balanced growth steady states". What is the key difference between the two?

I am guessing the difference is in terms of variable transformation in the model. In the case of the former, we don't need any stationary transformation, while in the case of the "balanced growth steady states", we do need the stationary transformation in the model.

But in both cases, the data is HP filtered while estimating. Is this correct? Any help will be great.

Some models have a unit root in the productivity process. If you have that, there isn't a single steady-state, but instead the real processes are cointegrated with a single common trend. The DSGE instruction is generally able to handle models like that, though you have to use the SOLVEBY=SVD option, which isolates the unit root. An example is included with RATS v8, which I've included below:

TomDoan wrote:Some models have a unit root in the productivity process. If you have that, there isn't a single steady-state, but instead the real processes are cointegrated with a single common trend. The DSGE instruction is generally able to handle models like that, though you have to use the SOLVEBY=SVD option, which isolates the unit root. An example is included with RATS v8, which I've included below:

dsgekpr.rpf

How can I use this model generate the data ? such as consumption , investment....

Thank U

iloverats wrote:

TomDoan wrote:Some models have a unit root in the productivity process. If you have that, there isn't a single steady-state, but instead the real processes are cointegrated with a single common trend. The DSGE instruction is generally able to handle models like that, though you have to use the SOLVEBY=SVD option, which isolates the unit root. An example is included with RATS v8, which I've included below:

The attachment dsgekpr.rpf is no longer available

How can I use this model generate the data ? such as consumption , investment....

Thank U

This is a different model, but the idea is the same. You simulate it using DLM(TYPE=SIMULATE) and invert the log-linearization to get the levels of the desired series.