Apologies for a very simple question.
Can I consider the long term average of GDP as steady state value? Or should I go for the model specific steady state value? Like find the steady state of an equation and then calculate it through # sign.
That depends on the context. But GDP is trending upwards. So the long-term average does not exist and therefore cannot be used. In contrast, using ratios often works.
What about using the FMOLS cointegration for prior from the data?
For example if I want to have steady state capital to net-worth (K/N) ratio, can I use the FMOLS for that?
Should I go for K as dependent and Net-worth as independent? Or the opposite?
Again, for a ratio, this should be fine. What do you mean with dependent/independent?
If one tries to find the data based value of prior, can the cointegration be used? If so, for example for K/N, should K (capital) be a dependent variable and Net-worth be an independent? Or vice versa?
- A prior must not be based on the data sample.
- If you are just interested in the ratio, you could simply use the average ratio. If you want to estimate cointegration, then K and N would both be endogenous in the vector error correction model.