Structural Changes in the Cointegrated Vector Autoregressive

This paper generalizes the cointegrated vector autoregressive model of Johansen (1988) to allow for structural changes. Estimation under various hypotheses is made possible by a new estimation technique, that makes it simple to derive a number of interesting likelihood ratio tests. E.g., the test for m structural changes against m+ k structural changes (occurring at fixed points in time), m ∈ N0, k ∈ N, and test of linear parameter restrictions when the null hypothesis allows for structural changes. The asymptotic distribution is χ in both cases. The model is applied to US term structure data, and structural changes in September 1979 and October 1982 — points in time with large changes in the Fed’s policy — are found to be significant. After accounting for these structural changes, I cannot, contrary to previous studies, reject the long-run implications of the expectations hypothesis.