This study was undertaken to apply the Empirical Bayes Estimation of Safety and Transportation (EBEST) methodology to actual HSIS data from Minnesota. Both EBEST and the "classical" before/after methodologies were used. A treatment group of 13 intersections where new traffic signals were installed and a reference group of 79 intersections were selected. After traffic signals were installed, the total number of crashes decreased 25 percent by the EBEST method and 30 percent by the classical method. Injury crashes decreased 23 percent by the EBEST method and 38 percent by the classical method. Thus, the classical method overestimated the treatment effect. More generally, a simple before-and-after comparison at a site where a safety treatment is selected based on its crash experience is likely to result in an overestimation of the treatments effect, because regression-to-the-mean is not accounted for.