Ripple effect of fundamental error in traditional factors

Abstract

The equal-weighted constraints of traditional factor model construction method adhere nonzero alpha problem. As the law of mathematics, persistent alphas consequently prevent factor portfolios to not span their optimal portfolio and their own basis portfolios which they are originated from because rank-deficient factor portfolios do not generate enough span space. Grinblatt and Saxena (2018) mainly suggested two solutions to improve traditional factor portfolios: 1) lift simple equal-weights constraints of traditional factor portfolio construction and 2) formulate a mean-variance benchmark (ex-post efficient combination) that is made with groups of sparsely sorted extreme quintile basis portfolios that mimic traditional factor portfolios. Their methodology has been implemented in the Korean stock market. Noted Gibbons, Ross, and Shanken (1989) GRS F-Score, Ledoit and Wolf (2008) LW circular bootstrap test results. The jackknife estimation has been employed to control in-sample optimism that may be prevailed in estimated sample statistics. Despite out-of-sample estimation handicap, the collective results indicate that jackknifed mean-variance optimized (MVE) benchmark corrects omitted price variable and tends to have higher pricing efficacy than their respective traditional factors across various settings. The suggested method corrects the omitted variable bias and empowers additional pricing ability to their origins.