I am estimating a rolling cointegration model and saving the parameters of the long run equation for trend analysis. I wonder if it is possible to estimate time varying parameter VECM using Kalman Filter on cointegrating equation instead of ecm or short run parameters. Many thanks mrrox

  • $\begingroup$ Is your question whether or not it is possible to cast the model in state space form and then let the long-run relations vector (often called the $\beta$ vector) vary according to a random walk process? $\endgroup$ – Plissken Oct 5 '15 at 8:49
  • $\begingroup$ Hello @ Plissken, yes, can I do that? if Yes, do you know of a code on the net or someone who could help with get such a specification estimated? $\endgroup$ – mr.rox Oct 5 '15 at 10:11
  • 1
    $\begingroup$ I'll write an answer up later on what you can do. Letting the long-run relations vector vary according to a random walk process is problematic but if you are willing to use a Bayesian framework it should be possible. See: personal.strath.ac.uk/gary.koop/kls8.pdf. Note that Koop also publishes his Matlab replication files on his website. $\endgroup$ – Plissken Oct 5 '15 at 10:31
  • $\begingroup$ Many thanks, it appears Koop published this paper with another guy in 2011. The code is in Gauss. Do you know if there is a matlab code of this sort? Thanks. $\endgroup$ – mr.rox Oct 5 '15 at 11:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.