I suppose another thing to consider from the academic standpoint is that when using y/y data for inflation (rather than q/q SA) is that one can get complicated MA structures in the errors. Wondering if these could result in biases in a model with lagged dependent variables.
Yes, that is right. Of course, even if one uses m/m data, the errors may have a MA(1) structure. Here the results and estimated dynamics appear broadly unaffected if a MA structure of the errors is allowed for. I am guessing that that is because the estimated MA parameters are “small.”
Interesting analysis. What measure of domestic inflation do you use?
If you go here:
https://data.snb.ch/en/topics/uvo/cube/plkopr
And look at “Item 2 of 15” there is a series called “domestic.”
The series is also available at the BFS’ website at https://www.bfs.admin.ch/asset/en/32267409
I suppose another thing to consider from the academic standpoint is that when using y/y data for inflation (rather than q/q SA) is that one can get complicated MA structures in the errors. Wondering if these could result in biases in a model with lagged dependent variables.
Yes, that is right. Of course, even if one uses m/m data, the errors may have a MA(1) structure. Here the results and estimated dynamics appear broadly unaffected if a MA structure of the errors is allowed for. I am guessing that that is because the estimated MA parameters are “small.”