Finite-sample properties of the Campbell and Thompson out-of-sample R2
研究了Campbell和Thompson样本外R²统计量在滚动和扩展窗口估计下的有限样本性质，发现该统计量存在向下偏差，弱预测下常出现负值，并通过模拟验证了近似精度，应用于股票溢价预测。
We study finite-sample properties of the Campbell and Thompson (2007) out-of-sample R 2 statistic ( R O O S 2 ). We derive analytical approximations for its expected value under rolling and expanding window estimation. R O O S 2 is downward biased relative to the true level of predictability, and negative values of R O O S 2 can arise frequently under weak predictability. Simulations confirm the accuracy of the approximations and describe the full distribution of R O O S 2 under different data generating processes. We discuss an empirical application to equity premium predictors.