Problems Related to Over-identifying Restrictions for Structural Vector Error Correction Models

Structural vector autoregressive (VAR) models are in frequent use for impulse response analysis. If cointegrated variables are involved, the corresponding vector error correction models offer a convenient framework for imposing structural long-run and short-run restrictions. Occasionally it is desirable to impose over-identifying restrictions in this context. Some related problems are pointed out. They result from the fact that the over-identifying restrictions have to be in the admissible parameter space which is not always obvious. Conditions are given that can help in avoiding the problems.