Traffic signals, if properly designed and justified according to current engineering practice, can improve the traffic safety and traffic operations at intersections. The effects of resulting crashes and the impacts on safety from the installation of traffic signals at unsignalized intersections are complicated, and the current literature and research on this topic are limited, even though traffic signals have been used for many years. The current crash warrants in the "Ontario Traffic Manual, Book 12: Traffic Signals" do not take into consideration the effect of traffic volume variations in crashes. It states that if conditions such as "five or more 'correctable' crashes per 12 month period averaged over a 36 month period" are satisfied, a signal may be warranted. The purpose of this paper is to develop and recommend a safety analysis and evaluation tool for estimating the expected safety of the installation of traffic signals through use of historical crash and volume data. The paper also provides a quantitative process tool for determining whether a traffic signal is justified at an unsignalized intersection. The proposed approach uses the empirical Bayes method and crash prediction models for estimating the safety effects on unsignalized intersections that are being considered for traffic signal installation. The performance of the proposed procedure is evaluated and analyzed on 17 unsignalized intersections taken from different jurisdictions in the province of Ontario, Canada.