Quasi-Experimental Design

A scientific experiment is a controlled set of observations aimed at testing whether two or more variables are causally related. William Shadish, Thomas Cook, and Donald Campbell describe two broad types of experiments: (a) randomized experiments, in which study units are randomly assigned to observational conditions; and (b) quasi-experiments, in which study units are not randomly assigned to observational conditions because of ethical or practical constraints. Although it is more difficult to draw causal inferences from quasi-experiments than from randomized experiments, careful planning of quasi-experiments can lead to designs that allow for strong causal inferences.

In order to infer a relationship between cause and effect, three requirements must be met: Cause must precede effect; cause must be related to effect; and, aside from the cause, no alternative explanation for the effect must be plausible. Randomized and quasi-experiments do not differ with respect to the first two requirements. However, with respect to the third requirement, randomized experiments have an advantage over quasi-experiments. Because study units are randomly assigned to conditions in randomized experiments, alternative explanations (e.g., confounding variables) are equally likely across these conditions and can be ruled out. But because quasi-experiments lack random assignment between conditions, alternative explanations are difficult to rule out. This entry focuses on the validity of, common designs of, and inferences drawn from quasi-experiments.