We use double-hurdle or Cragg models to test whether decisions about smoking participation and consumption levels are separate, endogenous choices. We also estimate several dominance models which assume that standard corner solutions are not applicable. A complete dominance model with independence between the participation and consumption decisions is the preferred specification for our data. The most important variables influencing participation are region, number of children, education, and ethnicity. Region, age, race, ethnicity, and health status most affected consumption levels.