The Experimentalist Looks Within: Toward an Understanding of Within-Subject Experimental Designs

The traditional approach in experimental economics is to use a between-subject design: the analyst places each unit in treatment or control simultaneously and recovers treatment effects via differencing conditional expectations. Within-subject designs represent a significant departure from this method, as the same unit is observed in both treatment and control conditions sequentially. While many might consider the choice straightforward (always opt for a between-subject design), given the distinct benefits of within-subject designs, I argue that researchers should meticulously weigh the advantages and disadvantages of each design type. In doing so, I propose a categorization for within-subject designs based on the plausibility of recovering an internally valid estimate. In one instance, which I denote as stealth designs, the analyst should unequivocally choose a within-subject design rather than a between-subject design.