An economic experiment places people in an environment desired by the experimenter, who then records the time paths of their economic behavior. Performing experiments using actual people at the level of national economies is obviously impractical but constructing a model economy and computing the economic behavior of the model's people is. Such experiments are termed ‘computational’ because economic behavior of the model's people is computed. This essay specifies the steps in designing a computational experiment to address some well-posed quantitative question. The computational experiment is an econometric tool used in the task of deriving the quantitative implications of theory.