Optimal Tests for Nested Model Selection with Underlying Parameter Instability

This paper develops optimal tests for model selection between two nested models in the presence of underlying parameter instability. These are joint tests for both parameter instability and a null hypothesis on (a subset of) the parameters. They modify the existing tests for parameter instability to allow the parameter vector to be unknown. It is commonly argued that out-of-sample rolling tests are useful to select between competing models when the parameters are time-varying. This paper argues that the optimal tests identified here are locally asymptotically more powerful than the out-of-sample rolling tests. It also shows that the optimal tests are more powerful than sequential tests that test for parameter instability in a first stage and select the model in a second state, the reason being that the two stages of the test are not independent. A simple empirical application to international finance models of nominal exchange rate determination is considered.