以折刀法減少估計式偏誤:方法與財務應用

Abstract

本篇論文介紹一種能減少估計式偏誤的方法: Quenouille (1956) 提出的折刀法 (jackknife)。我們也說明,在 GMM 的架構下,根據傳統折刀法所得到的估計式, 事實上與 Angrist, Imbens,and Krueger(1999)所介紹的折刀法工具變數估計式的作法並不相同。模擬結果顯示傳統折刀法的估計式與折刀法工具變數皆能皆能減少 GMM 估計式的偏誤,但傳統折刀法的估計式效果較好且適用範圍較為廣泛。實證結果亦支持傳統折刀法的估計式的確能減少 GMM 估計式的偏誤。

This paper introduces a bias reduction estimator method: Quenouille (1956) proposed jackknife. We also shows that in the GMM framework, according to the traditional jackknife estimator obtained, in fact, is not the same with Angrist, Imbens, and Krueger (1999) described jackknife instrumental variables estimation approach. Simulation results show that the traditional jackknife estimation method and jackknife instrumental variables estimator both can reduce the bias of GMM estimator, but the traditional jackknife estimator has better performance and wider application. The empirical results also support the traditional jackknife estimator can indeed reduce the bias of GMM estimator.