Following the paper https://www.jstor.org/stable/20111963?seq=1 I’m trying to estimate the output gap by using, among others, a time-varying Phillips curve.

The problem is that I get a smoothed estimate of beta that is a constant but I modeled beta as a random walk (the beta is the partial derivative of inflation with respect to the output). What am I doing wrong here?

data.xls (38.5 KB)

outputgap.mod (1.1 KB)

When using `mode_compute=5`

, I get a better mode but one parameter runs into the lower bound.

Thanks for the suggestions, I’ve incorporated these in the mod file. Alas I still get a constant beta, which is strange, given that I modeled it as a random walk. What other approaches can I take?

data.xls (38.5 KB)

outputgap.mod (1.3 KB)

Your model is nonlinear and you seem to be requesting a linearization around 0 for `ygap`

implying that `betaa*ygap`

will result in `betaa`

dropping out. In a nutshell `betaa`

is not identified.

This has helped me a lot. Thanks.