I am trying to replicate Agénor (2014): “Optimal Fiscal Management of Commodity Price Shocks”. I’ve replicated the same properties of the model as in the paper, but the resulting IRFs are different (see IRFs.pdf). for example, the consumption response to the oil price shock is in the form of a hump, which does not occur in my model. Also, when analyzing sensitivity, as in Agénor (2014), impulse responses do not change whatever the parameter value has changed. I attached the relevant files.
- Is the difference caused by a model error or in steady state calculation? I think the reason is a mistake in calculating the steady state. Is it calculation method correct?
- Does error in calculate the steady state result in model insensitivity to any changes in the value of a particular parameters?
Please, could you take a look at the relevant files and help me. And if someone has the original code of the model, please can you send it to me.
Any help would be much appreciated.
Thanks in advance.
agenor2014.mod (5.7 KB) IRFs.pdf (207.4 KB)
Professor @jpfeifer please could you take a look and help me.
Sorry, but I don’t have time to work on a full replication. There is nothing immediately suspicious here. I can only recommend checking all equations again. Generally, I would also recommend working with the nonlinear model. That is less error prone.
OK, thank you very much. I have a general question, what is the reason for model insensitivity when changing the values of some structural parameters, even though these parameters affect the dynamics of the model?
Usually I would say the reason is a coding error somewhere.
Your IRFs for taxes and government spending are very different from those you are trying to replicate. The lump sum tax magnitude is very large while the government spending is very small. Maybe check these and also anywhere that parameter appears first.
Thank you for your reply @bdombeck (Sorry, i saw it recently).
I did not understand what you mean by
Your earlier statements indicated the IRFs were robust to changing parameters which according to the original paper is not the case. So looking to the equations which include (or are supposed to include) the parameters you are expecting to have an impact might be a place to start looking for mistakes.