ARMA(1,1) in Dynare

Dear Johannes,

Could I ask you a question about ARMA(1,1) in dyanre?

I take log of oil price/GDP deflator data, and then use Hp-filter to obtain the cyclical part. Then I find this part data follow ARMA(1,1). Then is it right to write the oil price(which is log linearized in the model) in dynare as below:

P=aP(-1)+beP+c*eP(-1); and estimate a, b, c ,and the std of eP shock? If this is right, then I do not need estimate the constant term in the ARMA(1,1)?

Many thanks in advance!


  1. You are not supposed to estimate AR-processes on HP-filtered data. That filter is non-causal because it is two-sided.
  2. If you apply the HP-filter, the data is already demeaned. If the same is true for your model variable, you can use a process without a constant term.

I see. Many thanks! !!