IVX Tests for Return Predictability and the Initial Condition

We address the sensitivity of asset return predictability tests to the initial conditions of predictors. The IVX test of Kostakis et al. (2015, Review of Financial Studies) assumes asymptotically negligible initial conditions, which we show can result in large power losses for strongly persistent predictors. We propose a modified test that initialises the instruments at estimates of the predictors’ initial conditions, enhancing robustness and detection power. Additionally, a hybrid test is introduced, combining the strengths of the original and modified tests to deliver robust performance across varying magnitude initial conditions. Empirical and simulation results demonstrate the effectiveness of these approaches in improving predictability testing.