Using simulations, the paper shows that there is a trade-off in using CLS and 2SLS on the one hand and ML on the other when estimating the parameters of a bivariate threshold vector equilibrium correction model with regime-specific cointegration vectors.

Using simulations, the paper shows that there is a trade-off in using CLS and 2SLS on the one hand and ML on the other when estimating the parameters of a bivariate threshold vector equilibrium correction model with regime-specific cointegration vectors.