This article develops two block bootstrap-based panel predictability test procedures that are valid under very general conditions. Some of the allowable features include cross-sectional dependence, heterogeneous predictive slopes, persistent predictors, and complex error dynamics, including cross-unit endogeneity. While the first test procedure tests if there is any predictability at all, the second procedure determines the units for which predictability holds in case of a rejection by the first. A weak unit root framework is adopted to allow persistent predictors, and a novel theory is developed to establish asymptotic validity of the proposed bootstrap. Simulations are used to evaluate the performance of our tests in small samples, and their implementation is illustrated through an empirical application to stock returns.
- c15 - Statistical Simulation Methods: General
- c22 - "Single Equation Models; Single Variables: Time-Series Models; Dynamic Quantile Regressions; Dynamic Treatment Effect Models"
- c23 - "Single Equation Models; Single Variables: Models with Panel Data; Longitudinal Data; Spatial Time Series"
- g01 - Financial Crises
- g12 - "Asset Pricing; Trading volume; Bond Interest Rates"
- Block bootstrap
- panel data
- predictive regression
- sequential testing
- stock return predictability
- weak unit roots