 This simulation study compared three methods for estimating the size of key populations, KPs, affected by HIV, including men who have sex with men, female sex workers, and people who inject drugs. The study found that Bayesian model averaging of log linear models, LLMBMA, and Bayesian non-parametric latent class modeling, BLCM, were more accurate than information theoretic selection of log linear models, LLMAIC. Additionally, the study showed that two-list estimation is unnecessary and that three or more lists should be used when estimating the size of KPs. This article was authored by Steve Guttruder.