控制FDR和潛在因子下的共同基金表現

Abstract

本文在考慮潛在因子與多重檢定的問題下檢驗日本共同基金的績效。我們主要遵循 Giglio, Liao, and Xiu (2021) 的方法來辨識潛在的定價因子，處理基金報酬資料中的缺失值問題，且運用 screening Benjamini and Hochberg procedure 控制偽發現率(false discovery rate, FDR)。結果顯示，在這些基金中僅有 0.47% 在長期內被辨識為具有顯著績效。然而，我們也發現在短期內有較高比例的基金展現出顯著的績效，而這些基金在不同子期間的表現持續優於其他基金，但並未成為長期具有顯著績效的基金。最後，我們在不同的 FDR 水準下構建了由顯著績效基金組成的投資組合，這些投資組合的樣本外績效均優於日經 225，顯示出這些基金在樣本內的表現能成功轉化為樣本外收益，並帶來顯著的經濟價值。

This paper examines the performance of Japanese mutual funds while addressing latent factors and the issue of multiple testing. We follow the methodology of Giglio, Liao, and Xiu (2021) to identify latent factors, handle missing values, and apply the screening Benjamini and Hochberg procedure to control the false discovery rate (FDR). Among these funds, only 0.47% are identified as outperforming funds. However, a greater proportion of mutual funds demonstrate superior performance in the short term, which continues to outperform others across different subperiods, though their performance does not sustain over the long term. Finally, we construct portfolios of outperforming funds controlled at varying FDR levels, all of which outperform the Nikkei 225 out-of-sample, indicating that these in-sample alphas successfully translate to out-of-sample returns and generate significant economic values.