Binary Outcomes Aren't Independent—But How Do We Account for That?

Spatial/spatiotemporal interdependence matters in binary outcomes like civil war, but estimating it is hard. This paper explains the challenges and introduces a new method: maximum simulated likelihood via recursive importance sampling. It compares this approach to naive estimation strategies and shows how to calculate spatial effects on probabilities of outcomes.

New Method: Maximum Simulated Likelihood with Recursive Importance Sampling

• Addresses complex interdependence structures directly
• Provides consistent, asymptotically efficient estimates
• Overcomes limitations of traditional approaches

Civil War Application in Sub-Saharan Africa

• Demonstrates practical relevance for spatial econometrics
• Illustrates effects on conflict probabilities
• Shows how to interpret substantive implications from models