A quantitative theory of the HIV epidemic : education, risky sex and asymmetric learning

We explore learning about HIV infection odds from risky sex as a new mechanism explaining the Sub-Saharan Africa HIV epidemic. Our novel empirical evidence reveals a U-shaped relationship between education and being HIV positive across epidemic stages, which prompts the idea of asymmetric learning: more educated individuals potentially learn faster and update their (latent) beliefs about infection odds more accurately than less educated individuals, inducing earlier sexual behavioral change among the more educated. Our nonstationary model incorporates three HIV epidemic stages, chronologically: a myopic stage where agents are unaware of how risky sex causes infections, a learning stage where agents update beliefs on infection odds, and an ARV stage reflecting treatment introduction. Anchored in the micro evidence---explaining the HIV-education gradient---we find that our learning mechanism is powerful: a 5-year earlier learning reduces new AIDS deaths by almost 45%, and a 10-year earlier learning results in a 60% drop.