Selective Conscription: Sequential Allocation under Incomplete Information in Brazil's Military Recruitment and Retention System

The value of military service is often learned only through experience, yet initial selection into service must occur before that experience takes place. This creates a tension at the heart of every selective conscription system: the military must allocate positions among a large pool of potential recruits under imperfect information about both aptitude and preferences. Brazil, which registers over 1.5 million young men each year but incorporates only about 5 percent, is one of the largest selective systems in the world. Yet no formal framework models this process as a sequential allocation mechanism. This paper develops such a model, combining centralized screening, capacity constrained selection by military organizations, and post-service learning through reenlistment. Two results follow. First, the interaction between screening and capacity constraints endogenously generates local quasi-random assignment at the selection margin, providing an institutional micro-foundation for regression-discontinuity identification without requiring an explicit lottery. Second, when preferences are learned through experience, conditioning initial selection on stated willingness to serve is suboptimal for the formation of a high-quality retained force. As hybrid voluntary-mandatory designs return to the European policy agenda, these results offer analytical building blocks for the reform of conscription systems under uncertainty.