Finding a suitable alternative for OLS

[thomaskm]

Dear forum participants.

For my bachelor thesis i'm researching inflation persistence. I am currently struggling with finding a suitable model for the corresponding regression specification: inflation=c+inflation(-1).
Since inflation is a non stationary process, OLS is unsuitable, unless i transform the series. Logarithmic transformation is not possible, since in some cases there is deflation (negative inflation).
Taking the first difference is not suitable either, since d(inflation)=c+d(inflation(-1)) no longer tests inflation persistence.

Any advice on how to resolve this issue would be greatly appreciated,

Thanks in advance,

Thomas.

[thomaskm]

20 views and no response yet, someone has to have an idea?

[startz]

How do you define persistence?

[thomaskm]

Persistence in this case means that inflation is dependent on its past value.
so : inflation = c+b1inflation(-1)+et. If b1 turns out to have a high value, i say that inflation is very persistent.
Hope you can give me some more advice on the basis of this information :)

[startz]

If you do it in first differences, then inflation is still dependent on past values--just in a more complicated way.

In fact, if inflation is nonstationary then shocks persist forever.

You might want to look up "impulse response function"

[thomaskm]

Thank your for this insightful comment. My supervisor told me that d(inflation)=c+d(inflation-1) would not measure persistence, but i think, like you, that he's probably wrong on this part. If inflation at time t increased with 2 percent, and at time t+1 with 1.8 percent, then inflation finds it hard to get back to a 0 (or negative increase). Yes, i think this reasoning is logical, lets hope my supervisor agrees. Thanks