Portfolio selection with robust estimators considering behavioral biases in a causal network

Abstract

In this study, we develop a behavioral portfolio selection model that incorporates robust estimators for model inputs in order to reduce the need to change the portfolio over consecutive periods. It also includes Conditional Value at Risk as a sub-additive risk measure, which is preferable in behavioral portfolio selection. Finally, we model a varying risk attitude in a causal network in which investor behavioral biases and latest realized return are related to using a causation algorithm. We also provide a case study in Tehran Stock Exchange, where the results disclose that albeit our model is not mean-variance efficient, it selects portfolios that are robust, well diversified, and have less utility loss compared to a well-known behavioral portfolio model.