Hi there! I am actually trying to simulate a small New Keynesian Model with constant gain learning. I wrote the code and in the IRF that I obtain, it is possible to see the persistence that the literature talk about. However, reading the doctoral thesis of Sergio Santoro (“Learning in Monetary Economics”) , in page 42 he assumes that agents believe that inflation and output are continuous invariant functions of the cost-push shock only, and given the **iid** nature that he assumes of the shocks, this implies that the conditional and unconditional expectations of inflation and output gap coincide. Thus, agents estimate them using simple means. So, then he writes the learning algorithm for each variable.

In my code, I wrote the same, but I am assuming AR(1) shocks. Then, is it correct to use that learning algorithm in the same way as he did?

My code:ALNKM_mcshock.mod (2.4 KB)

Sergio Santoro’s doctoral thesis https://www.tesisenred.net/bitstream/handle/10803/7351/tss.pdf?sequence=1

Thank you very much!