This article describes a risk analysis methodology that uses, in one hand, empiric estimations of recurrence based on information available in local disaster data bases, with which it is possible to estimate the loss recurrence of small events, and on the other hand, probabilistic analytical evaluations to estimate the loss recurrence of greater or catastrophic events for which there is not information due the lack of historical data. Using a “hybrid” loss exceedance curve it is possible to represent the disaster risk in a proper and complete way. The first segment of the curve that corresponds to the small and moderate losses, is calculated using an inductive and retrospective analysis, and the second segment of the curve that corresponds to the extreme losses, is calculated using a deductive and predicted analysis, in prospective. Two case studies are presented (Colombia and Mexico) where this new probabilistic risk assessment technique has been applied.