Spatially randomized designs can enhance policy evaluation

This article studies the benefits of using spatially randomized experimen tal designs which partition the experimental area into non-overlapping units with treatments assigned randomly to these units. Such designs improve pol icy evaluation in online experiments by providing more precise policy value estimators and more effective testing algorithms than traditional global de signs, which apply the same treatment across all units simultaneously. We examine both parametric and nonparametric methods for estimating and in ferring policy values based on the spatially randomized designs. Our analysis includes evaluating the mean squared error of the treatment effect estima tor and the statistical power of the associated tests. Additionally, we extend our findings to the dynamic setting with spatio-temporal dependencies, where treatments are allocated sequentially over time, and account for potential tem poral carryover effects. Our theoretical insights are supported by comprehen sive numerical experiments.