In this article, the authors discuss program evaluation of intervention studies when the outcome of interest is collected routinely at equally spaced intervals of time. They illustrate concepts using data from the American Stop Smoking Intervention Study, where the outcome is state per capita tobacco consumption. States differ widely in mean tobacco consumption, and these differences should be accounted for in the analysis. A large difference in the variance of the intervention effect maybe obtained depending on whether the variation in the between-state effects are considered. The confidence limits obtained by ignoring between-state effects are too optimistic in many cases.