The Markovian approach to modeling prevalence can be distinguished from purely statistical techniques such as multiple capture, Poisson estimation, and synthetic estimation and from such elaborate behavioral models as the system dynamics model. The Markovian model incorporates one or more States and transition parameters that determine flows between those States. It separates cocaine users into two categories, light users and heavy users, and demonstrates that the fraction of heavy cocaine users has varied greatly over time. The authors show that the effect of government programs designed to reduce heavy cocaine use will only be realized many years later and that the effectiveness of local law enforcement programs will also be delayed. The fact that various control programs focus on different aspects of cocaine use means that some strategies may be most appropriate for specific stages of the epidemic. The increased prevalence of heavy versus light cocaine users and its effect on total cocaine consumption is discussed. 21 references, 8 tables, and 42 figures