发表机构
Carl H. Lindner College of Business, University of Cincinnati; Católica Lisbon School of Business and Economics, Universidade Católica Portuguesa; Department of Statistics, Pennsylvania State University（辛辛那提大学卡尔·H·林德纳商学院; 葡萄牙天主教大学里斯本工商经济学院; 宾夕法尼亚州立大学统计系）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究探究良性过拟合是否适用于股权收益预测，发现无岭回归模型预测风险呈双下降模式，最优岭模型虽更优但差距随参数规模增大可忽略，且两类模型均未优于历史均值，证实标准股权预测因子缺乏真实预测能力。
AI 中文摘要
高参数化模型在复杂领域插值训练数据时仍能实现良好预测，这对经典的偏差-方差权衡提出了挑战。我们研究这种“良性过拟合”现象是否延伸至股权收益预测领域。与近期统计理论一致，我们记录了两个关键现象：其一，无岭回归模型的预测风险呈现双下降模式；其二，尽管最优岭回归模型始终优于其无岭对应模型，但在参数与观测值的比例较大时，该性能差距可忽略不计。然而，最终两种模型均未能优于简单的历史均值。这一实证证据与斜率系数为零的零假设下的渐近结果相符，表明即使在高度灵活的非线性机器学习架构中，标准股权预测因子也缺乏真实预测能力。这些发现协调了资产定价领域现代与经典机器学习的关系：在无真实信号的情况下，它们渐近收敛至历史均值基准。
英文摘要
Highly overparameterized models often predict well despite interpolating training data in complex domains, challenging the classical bias--variance tradeoff. We investigate whether this ``benign overfitting'' phenomenon extends to equity return prediction. Consistent with recent statistical theory, we document two key phenomena: first, a double descent pattern in the ridgeless model's prediction risk; and second, that while the optimal ridge model consistently outperforms its ridgeless counterpart, this performance gap becomes negligible at large parameter-to-observation ratios. Ultimately, however, both models fail to outperform a simple historical average. This empirical evidence aligns with our asymptotic results under the null hypothesis of zero slope coefficients, suggesting that standard equity predictors lack true forecasting power---even within highly flexible, nonlinear machine learning architectures. These findings reconcile modern and classical machine learning in asset pricing: in the absence of a true signal, they asymptotically collapse to the historical average benchmark.
