Freeway collisions are thought to be affected by traffic states. To reduce the number of collisions, the study to reveal how the traffic states influence collisions are required. Therefore, the purpose of the paper is to suggest a method to relate traffic states to collision frequency in freeway. We first defined section- based traffic phases showing traffic state of a section using upstream and downstream traffic states: free flow (FF), back of queue (BQ), bottleneck front (BN) and congestion (CT). Secondly, by integrating freeway collision data and traffic data from the California PeMS database, over a three-year period, we obtained the collision frequency for each traffic phase, and compared for a 32mile section of the I-880 freeway. The results show that collision rate in BN, BQ, and CT phase are approximately 5 times higher than the collision rate in FF. Also, the proposed method shows potential for predicting collision frequencies on freeway sections when combined by traffic simulation.