Bootstrap testing for nonlinear adjustments in error-correction models of nonsmooth transition vector

Nonlinear adjustment tests are presented for the error-correction models of two types of transition vectors with null hypothesis of linear and nonsmooth transition relationship The SupWald type test is adopted for testing the nonlinear adjustment, and the null asymptotic distribution of which is derived. The approach for simulating the p-values is then given. A residual bootstrap approximation is also presented. Finally, the simulation shows the desired performance with finite samples, and specifies the range of application. Applying the proposed methods to the U.S. treasury, we found the strong evidence for nonlinear adjustment.