A new safety indicator takes into consideration the risk-taking behavior of drivers as well as the prevailing traffic conditions at an intersection. The indicator is based on the idea that an intersection at which drivers are willing to take a higher risk is not as safe as one at which drivers are not willing to take high risks. Driver risk-taking behavior is modeled as a function of a driver's reaction to a possible collision scenario. Binary logistic regression was used to understand how the probability of a driver reacting to a possible collision scenario changes as a function of the variables defining the scenario. The data collection and safety index definition are presented from the perspective of permissive left turns; however, the concept of risk taking is universal; thus it is a feasible alternative for other maneuver types if appropriate data are obtained. Use of a safety index based on risk taking helps solve the engineer's dilemma of which of two intersections that have no crash history, or that have equal crash history, should be the target of a safety improvement program. The methodology presented can remove the subjective judgment that often takes place in such a scenario and provides the engineer with an objective alternative.