Using subspace algorithm cointegration analysis: Simulation performance and application to the term structure

The performance of subspace algorithm cointegration analysis is discussed in detail by means of both a simulation study, as well as an empirical investigation of the expectations hypothesis of the term structure on four interest rate and bond yield data sets. A new better performing order estimation criterion is introduced. The main finding of the simulation study is that the methods offer some advantages in determining the cointegrating rank, where they often outperform Johansen cointegration analysis. In the empirical application we find much stronger evidence for the cointegration implications of the expectations hypothesis of the term structure than when using VAR cointegration analysis. These favorable findings highlight the potential usefulness of the computationally simple yet widely applicable method.