Beliefs and Portfolios: Causal Evidence

This paper examines how individuals form expectations about asset prices, addressing the shortcomings of traditional models that rely on the assumption of full information rational expectations (FIRE). The authors use a randomized information experiments to test competing theories of expectation formation and their implications for asset pricing. The study’s findings highlight belief overreaction and subjective mental models as critical components in understanding how individuals process information and make investment decisions.

The authors’ randomized experiments provide causal evidence on how different types of information, such as past returns, earnings growth, P/E ratios, and expert forecasts, affect beliefs about future returns. By comparing treated individuals with a control group, the authors document significant deviations from rational expectation benchmarks. They show that belief overreaction dominates models of underreaction, such as rational inattention or noisy information. Moreover, the findings indicate that individuals’ reactions to information are influenced by their perceived importance of the data, even when they are relatively well-informed about it.

Importantly, the authors find that belief overreaction is a central feature of how individuals update their expectations. Participants in the study tended to excessively extrapolate past stock returns and earnings growth into their forecasts of future returns, even when objective evidence suggested these processes were independent. This overreaction is inconsistent with models that assume rational or modestly imperfect information processing. A particularly striking finding relates to individuals’ responses to price-earnings (P/E) ratios. While academic literature has shown an inverse relationship between P/E ratios and future returns, participants reacted to higher P/E ratios by increasing their expected returns, suggesting a fundamental misunderstanding of the predictive power of valuation metrics.

The study identifies two key drivers behind these biases: limited information about economic variables, such as P/E ratios and past returns, and flawed subjective models of how the economy functions. These subjective models, which individuals rely on to interpret and predict economic dynamics, differ significantly from the objective relationships established by empirical data and theory. By linking individuals’ prior perceptions to their expectations, the authors demonstrate how these subjective models shape the way participants process new information.

Despite these belief frictions, the study finds that individuals’ investment decisions are broadly consistent with the Merton model of portfolio choice when conditioned on their expectations. For example, the sensitivity of individuals’ allocation to risky assets relative to their expectations aligns with standard economic predictions. However, this alignment is only achieved after accounting for the exogenous variation in beliefs generated by the experimental treatments, highlighting the importance of separating belief formation from decision-making behavior.

In conclusion, this paper contributes to the asset pricing and household finance literature by providing robust empirical evidence for heterogeneous and subjective mental models in belief formation. It underscores the need for theoretical models that account for such heterogeneity to better understand asset pricing dynamics, trading volume, and wealth redistribution. By highlighting the role of frictions in information processing and subjective belief systems, the study opens new avenues for research into the mechanisms driving individual and aggregate financial behavior.