One emerging type of ad-hoc network is the Vehicular Ad-Hoc Network (VANET), in which vehicles constitute the mobile nodes in the network. Due to the prohibitive cost of deploying and implementing such a system in real world, most research in VANET relies on simulations for evaluation. A key component for VANET simulations is a realistic radio propagation model that ensures conclusions drawn from simulation experiments will carry through to real deployments. Prior work has shown that the idealized perfect-within-range models commonly used in network simulation tools can be misleading. In this paper, we set out to study the effect of radio models on the simulations of vehicle communication. We first conduct measurements of the received signal strength between moving vehicles. The data obtained from these experiments are then used to develop an empirical radio propagation model. Finally, we evaluate the effect of our trace-driven model on several routing protocols in the popular network simulator NS2.