This paper proposes a method for identifying event sequences leading to blackouts in power systems. To identify the cascades, grid is modeled like an event tree where each branch represents a contingency and each node represents state of the system. Due to huge number of possible paths in the event tree, computational burden is the key issue in searching the cascades. In this paper, computational burden has been reduced in two ways. In the first step, search space is reduced by defining a region of vulnerability and in the second step, less important events in the region of vulnerability are discarded using a risk index which condenses both the severity and probability of any event. Proposed method is tested on benchmark New England 39-bus system and simulation results are presented to illustrate the effectiveness of the method.