We present a theoretical framework that provides a logical structure for information collection, information analysis, and decision making in a vehicular ad hoc network (VANET) environment. Our framework begins by describing a threat model. Then we introduce mechanisms for excluding malicious vehicles and messages from corrupting the system. We introduce a set of algorithms to build a reference list from the crowd based on ground truths. Vehicles compare messages received from multiple agents against a reference list stored in the road side unit (RSU). As a complement, vehicles that are not in proximity to RSUs aggregate opinion from neighbouring vehicles to establish the trustworthiness of a message. A vehicle makes a decision on received messages upon cross validation, and forms a posterior belief about an agent. We apply conditional probability to check for the accuracy of messages. Our framework is dynamic and scalable in that the reference list periodically updates itself to account for elapsed time and frequently changing road conditions. The framework provides a universal template for modelling trust management in VANET.