A multivariate innovations state space Beveridge-Nelson decomposition

 

The Beveridge-Nelson vector innovations structural time series framework is a new formulation that decomposes a set of variables into their permanent and transitory components. The proposed framework is flexible, modelling inter-series relationships and common features in a simple manner. In particular, it is shown that this new specification is simpler than conventional state space and cointegration approaches. The approach is illustrated using a trivariate data set comprising the GDP of Australia, the USA and the UK.

Keywords: vector innovations structural time series model, multivariate time series model, Beveridge-Nelson decomposition, common components.

Cite

@article{dSHS09,
  author    = {Ashton {de Silva} and Rob J Hyndman and Ralph D Snyder},
  doi       = {10.1016/j.econmod.2009.04.004},
  journal   = {Economic Modelling},
  number    = {5},
  pages     = {1067--1074},
  publisher = {Elsevier},
  title     = {A multivariate innovations state space {Beveridge-Nelson} decomposition},
  volume    = {26},
  year      = 2009
}