The document compares three methods for hotspot identification: crash rate, empirical Bayes, and Bayesian hierarchical models. It evaluates the performance of the methods using four tests across different time periods on crash data from intersections in Germany. The Bayesian hierarchical model outperformed the empirical Bayes approach on most tests and time periods according to the evaluation scores. Crash rate was the least recommended method. The study concludes the Bayesian hierarchical model is superior but notes limitations from a limited number of variables and comparing crash predictions rather than observed crashes.