Belief Updating and Investment Decisions: The Impact of Good or Bad News Varies With Prior Returns

Abstract

Combining prior findings on belief formation implies that investors' belief updating depends on an interaction between information favorability and the sign of the investment's return.

In a pre-registered experiment we observe the hypothesized expectation dynamics, leading to known profit harming trading behaviors.

A context sensitive Reinforcement Learning model describes this expectation formation best and its complexity is justified by the data.

As the main driver of the interaction we observe that learning from unfavorable information is stronger for gain compared to loss positions.

Providing participants with additional information about the price movements mitigates these effects.

Keywords: beliefs, expectations, reinforcement learning, information, disposition effect, investing, cognitive model

JEL Classification: D01, D12, D83, D84, G11, G41