Localization systems play a major role in many applications for Vehicular Ad Hoc Networks (VANets). Although Data Fusion techniques can provide reliable localization information for most of the application requirements in VANets, enhancements on the localization systems are required and desirable. Unique characteristics of VANets such as mobility constraints, driver behavior, and high speed displacement nature of vehicles cause rapid and constant changes in the network topology, leading to the dissemination of outdated localization information. To circumvent this problem, an alternative is the use of predicted future locations of vehicles. The main idea of this approach is to use the localization prediction as an extension of a Data Fusion localization system. In such approach, a future position of a car is predicted for a given future time step and used to take advantage of a future time-space window of a vectorial trajectory rather than a static localization point. Thus, in this paper we further discuss this subject by analyzing the use of localization prediction as natural way to improve applications for VANets. We present a set of experiments that shows the results of such techniques when applied to a realistic VANet scenario.