Page 24 of the secret sauce says: For an AR(1) model to be covariance stationary: -b1 must be less than 0 -if b1=0, there is a unit root -The mean reverting level must be defined The first two are throwing me off…I thought if b1=1 then b0/(1-b1) is undefined and therefore unit root. Where have I gone wrong?
I think you are confusing two things: the mean reversion formula b0/1-b1 and the Dickey Fuller test. Dickey fuller tests the Ho: that g=0, where g= 1-b1. To confirm, if b1=1, you have a unit root, undefined mean reverting level and are not covariance staionary. If you reject the null in your Dickey Fuller test then g is not 0, therefore b1 is not 1.
Yeah… Thats how the secret sauce has it. It has to be an obvious mistake.