2025-05-20-欧洲央行-随机对照试验中的理性疏忽（英）_32页_1mb

Rational Inattention during an RCT Summary

We introduce an information provision experiment into a dynamic rational inattention (RI) model and analyze how the treatment effect varies with environmental and individual characteristics. The key findings are as follows:

1. Treatment Effect Variation:

   • The treatment effect is strictly decreasing in the importance of being informed (γ) and the variance of the state innovation (σ²) in the standard case, where agents allocate attention across periods.
   • The effect is increasing in the marginal cost of attention (µ) for a given ratio of costs during the survey (λ/µ).

2. Corner Solutions:

   • In cases where agents pay no attention outside the survey, the treatment effect is increasing in γ.
   • The nature of the provided information (previously available or not) determines whether the treatment effect increases or decreases with γ.

3. Interpretation of Empirical Findings:

   • The model explains why treatment effects are smaller during high inflation (due to increased γ and σ²).
   • Individuals with lower γ or higher µ show smaller treatment effects, aligning with ties to monetary policy importance.

4. Policy Implications:

   • Communication restating existing information has less effect on those who care more about being informed when information is more important.
   • Expanding publicly available information may be more effective for informed individuals at crucial times.

5. Broader Scope:

   • The model applies to various experimental designs, including forecasts or noisy state measures.
   • Additional monetary incentives during surveys increase attention but do not alter the γ-dependence of the treatment effect.

This analysis provides a benchmark for interpreting real-world information provision experiments under rational inattention.