The RATS Software Forum

Dear Tom, I am seeking for your help again. I would like to know whether it is possible to use mean-switching in msregression. I am trying detect partisan cycle in US economic growth. The standard test to the existence of political business cycle is to run an auto-regression of an economic measure on itself (such as gdp growth), a small set of economic variables, and political dummies. My primitive model to test the partisanship effect on economic growth is to regress growth rate on its own lags (4 lags) and party dummy. I would like to incorporate mean-switching process (Hamilton) into the model and see how it will affect the party dummy after taking care of the no-linearity in the U.S. macro data . Thank you in advance

In general, no. The AR and VAR can do mean switching because they are self-contained. As soon as you start adding dummies (or any other variable that isn't a lagged dependent variable), the mean of the process changes with those.

TomDoan wrote:In general, no. The AR and VAR can do mean switching because they are self-contained. As soon as you start adding dummies (or any other variable that isn't a lagged dependent variable), the mean of the process changes with those.

Thank you for the reply. so is it right to switching the intercept as an alternative way ?

Yes.