A simulation analysis of conditional tests for parameter stability in cointegrated VAR models

This paper conducts a simulation study of parameter stability tests using conditional cointegrated vector autoregressive (CVAR) models. Monte Carlo simulation experiments show that, under the assumption of weak exogeneity, the parameter stability tests based on conditional CVAR models are more powerful than those based on a joint CVAR model. However, the reverse is observed in the experiments when the assumption fails to hold true. The overall assessment of the simulation study leads to the formulation of a practical procedure for testing the constancy of parameters in both conditional and joint CVAR models.