A Comment on Tests for Asymmetric Threshold Cointegration with an Application to the Term Structure: Cointegration Methods Matter

Abstract

Wohar and Sollis (2007) test for asymmetric threshold cointegration in theterm structure of interest rates for several OECD countries. They use the Engle-Granger(1987) method to construct the error correction term, and this estimated error correctionterm is then used to test for asymmetric threshold cointegration. The Engle-Grangermethod is commonly used in the term structure and pass-through literature. However, weshow that different cointegration tests and methods may potentially lead to different errorcorrection terms. As noted by Cook (2008), the Engle-Granger method leads to severedownward bias in the long-run (cointegration) parameter. These parameters are then usedin the second stage to test for asymmetries. We test for cointegration using theEngle-Granger and the Johansen (1991) methods. Using the same sample period for asubset of countries used in Wohar and Sollis, our results suggest substantial downwardbias in the Engle-Granger method, which may potentially alter any asymmetric adjustmentand/or threshold cointegration results (C22, E43).