Detecting multiple structural breaks on a Koyck distributed lag model: A Bayesian approach

Abstract

Koyck Distributed Lag Model undergoing a possible multiple structural changes occurring at unknown positions in time is analyzed using Bayesian approach. The posterior distribution of the number of change points and the positions of the structural change points are derived using filtering recursions similar to forward-backward algorithms. The segments are assumed to be independent, and so are the priors. Normal and gamma priors are used for lag weights and variance, respectively. The priors used for number of change points and position of change points are binomial and uniform discrete distributions, respectively. Only the parameter β’s (lag weights) are allowed to change.