Culture and Coordination Failures: Essays on Depositor Heterogeneity in Financial Panics

Walter Stover

Advisor: Thomas Stratmann, PhD, Department of Economics

Committee Members: Jonathan Schulz, Robert Axtell

Carow Hall, #1
November 19, 2026, 01:00 PM to 02:00 PM

Abstract:

This dissertation examines the role of individualism as a psychological characteristic partially shaped by the cultural background of depositors in financial panics. Using empirical data from a run on a cryptocurrency platform, it shows that depositors' psychological backgrounds are associated with heterogeneous behavior in runs, consistent with predictions from a global games model with signal weighting.

Chapter 1 develops a theoretical framework grounded in the global games model, extending it to incorporate heterogeneous signal weighting. Individualism is modeled as affecting the relative weight a depositor places on private versus public signals. This framework allows for persistent partial runs, conditional on a stale public signal and stable dispositional weights: when public information is more optimistic than fundamentals, more individualistic depositors run at fundamentals where more collectivist depositors stay, and vice versa. This chapter also finds that the individualists-only range of runs widens with public error and the weight-odds gap: more reassurance means more sorting. Finally, individualism shows up only under disagreement about the signals: when public and private information agree, type differences shrink to noise.

Chapter 2 examines the empirical results of investigating the role of individualism in users during a cryptocurrency bank run using a novel dataset for a run that occurred on the cryptocurrency platform Celsius. We use the global games theoretical framework developed in Chapter 1 to make several predictions about how we expect individualism, as measured through name-based proxies, to affect user behavior during this cryptocurrency bank run. First, that individualists withdraw more and earlier; second, that the individualism gradient strengthens as the platform's official narrative grows more stale; third, that the individualism gradient is not as prominent in times when the official narrative is closer to the fundamentals; and fourth, that individualists react more to platform-specific signals on fundamentals, and less to common focal points. Our findings from analyzing the Celsius empirical data are consistent with all four hypotheses.

Chapter 3 presents a complementary agent-based model (ABM) that investigates the dynamics of sorting by individualism in conditions that Chapter 1's framework cannot make predictions in. This ABM simulates heterogeneous depositors embedded in a Newman-Watts small-world network who make withdrawal decisions via Monte Carlo inference about the platform's solvency. Culture endows depositors with individualism endogenously through a Flache-Macy warm-up phase, forming the dispositional weights depositors place on socially transmitted signals versus their own private assessment of fundamentals. In this chapter, agents have local social signals in the form of neighbors' withdrawals, which are pessimistic during a panic event. Consistent with Chapter 1's framework, this chapter finds that collectivists run first, while individualists stay: the converse example to Chapter 2 where the public signal was more optimistic and individualists ran first. This chapter also finds evidence for the effect of cultural share in the population, with the run probability falling monotonically with a larger share of more individualist depositors. However, we do not find evidence that mixed populations are inherently more fragile because of individualist-seeded information cascades.