Reading 13 - time series analysis: stationarity and mean reversion

CFA Level II
#1

2 rather complicated doubts

On page 243 of schweser…besides the 5 possible results we obtain from testing whether 2 time series are covariance stationary, what if one of the series is nonstationary BUT the 2 series are co integrated…why cant linear regression be used in that case?

Also regarding mean reverting levels and unit roots…on page 230 (last paragraph) what if b1>1 then we would still have a finite mean reverting level in that case right? why cant the coefficient be greater than 1?

thanks

#2

  1. Not sure if what you’re asking/thinking makes total sense. If one of the series is nonstationary, your regression model already has a problem. You shouldn’t even be using the series.

  2. The coefficient can be greater than one. It’s just that it is an advanced topic, and the curriculum doesn’t go into those details.

Hope this helps.