Risky Investments under Static and Dynamic Information Acquisition

Abstract

This paper examines the trade-offs between static and dynamic information acquisition in risky investments, with a focus on clinical trials. In a stylized model, a risk-averse decision maker with a Constant Absolute Risk Aversion (CARA) utility function gathers costly data before choosing an investment level. Under a static approach, information acquisition is fixed in advance; under a dynamic approach, data collection can continue or stop based on interim outcomes. Our analysis reveals three main findings. First, falling marginal costs of information lead to increased data collection, which raises average investment but can also amplify losses in adverse states. Second, allowing for dynamic sampling typically results in higher expected information acquisition and investment than a static design, especially when prior beliefs about the investment’s returns are neither too optimistic nor too pessimistic. Finally, somewhat counterintuitively, less risk-averse decision makers acquire more information because they invest more in the underlying project, making refined forecasts more valuable. These insights highlights the managerial advantages of flexible sampling schemes in settings like vaccine development, while offering guidance for policymakers regulating data collection in high-stakes industries.

Keywords: Information acquisition; Investment; Bayesian learning;, Bayesian learning, Investment

JEL Classification: D81, D83, G11