2026 Theses Doctoral

Three Essays on Information: Learning, Fragility, and Privacy

Liu, Tianhao

This dissertation studies how information shapes economic outcomes across three distinct but connected environments.

The first chapter analyzes sequential social learning and shows that asymptotic learning depends on a joint property of preferences and information rather than on information alone. It develops the condition of excludability and uses it to explain why learning can succeed under familiar information structures that fail standard unbounded-beliefs conditions in multi-state settings.

The second chapter studies statistical learning under misspecification and shows that learning is not uniformly robust, even in a simple passive setting. Although misspecification may be arbitrarily small, long-run beliefs can still be severely distorted if the signals are sufficiently uninformative. This has implications for information design and persuasion.

The third chapter develops an axiomatic framework for privacy measures grounded in a worst-case principle of uniform protection. It characterizes worst-case privacy measures, studies their structure, and applies them to matching and voting environments to show how informational value and privacy protection trade off in institutional design.

Taken together, the three chapters examine the power, fragility, and limits of information. Across learning, inference, and privacy, they show that economic analysis must account not only for how much information is available, but also for how it is structured, how it is interpreted, and when it should be constrained.