The Paradox of High-level Military Dismissals in China: A Formal Bayesian Model

Abstract Between 2023 and 2024, China dismissed dozens of senior military officers, including two defense ministers personally appointed by Xi Jinping. These purges occurred not during a political crisis but at the peak of an ambitious modernization campaign. Why? We develop a formal model showing that high-level dismissals are an equilibrium feature of centralized military modernization under imperfect monitoring, not evidence of regime weakness or factional struggle. Our key insight: when observable signals conflate corruption with professional deviations from centralized blueprints, rational principals cannot distinguish honest implementation disagreements from rent-seeking. We prove a Variance Compression Lemma: under noisy monitoring, dismissals optimally compress variance around the blueprint, even when this eliminates competent officers. The model predicts that dismissals cluster after doctrinal reforms rather than during political crises, concentrate in services with complex modernization programs, and exhibit high stochastic variance—officers with identical behavior face divergent fates due to signal noise.