Visual Inference and Graphical Representation in Regression Discontinuity Designs

Despite the widespread use of graphs in empirical research, little is known about readers' ability to process statistical information they are meant convey ("visual inference"). We study visual inference within context regression discontinuity (RD) designs by measuring how accurately readers identify discontinuities produced from data generating processes calibrated on 11 published papers leading economics journals. First, we assess effects different graphical representation methods using randomized experiments. find that bin widths and fit lines have largest impacts whether participants correctly perceive presence or absence a discontinuity. Incorporating experimental results into two decision theoretical criteria adapted recent literature, small bins with no construct RD performs well recommend it as starting point practitioners. Second, compare widely used econometric procedures. achieves similar lower type I error rates complements inference.