Signaling games help explain political crises, but they often have multiple possible outcomes. This creates a problem for using theory with data.

The Problem:

Multiple equilibria make likelihood calculations unclear and discontinuous even if the game itself has one clear outcome.

Our Experiments Show:

Current estimation methods fail to find correct parameters in crisis-signaling games, despite large samples or unique equilibria.

Proposed Solutions: Three New Estimators

These new approaches overcome these limitations and perform better than existing ones. We tested them on data about economic sanctions.

Crucial Findings: Our analysis reveals a previously unnoticed U-shaped relationship between audience costs and leaders' willingness to threaten sanctions, showing the complex reality that simpler models miss.