CMStatistics 2022: Start Registration
View Submission - CFE
Title: Shrinking against sentiment: Exploiting behavioral biases in portfolio optimization Authors:  Nathan Lassance - UCLouvain (Belgium) [presenting]
Alberto Martin-Utrera - Iowa State University (United States)
Abstract: The performance of mean-variance portfolios is shown to be the sum of a market and an arbitrage component and the exposure of a mean-variance portfolio to each component is shown to depend on their in-sample performance. Consequently, mean-variance portfolios are highly affected by the arbitrage component and suffer from large estimation errors. However, shrinking the sample covariance matrix of returns toward the identity allows mean-variance portfolios to give more relevance to the market and alleviate the impact of parameter uncertainty. We time the exposure to each component by shrinking more when investor sentiment is low, which provides sizable economic gains with lower turnover than competing benchmarks.