Aiming at explaining the time-path of presidential approval and other binary popularity measures we propose a dynamic epidemiological model. In this approach the characteristics of communication-processes in heterogeneous electorates are formulated by help of differential equation systems reflecting the interaction between different groups of voters. The models' parameters can be used to cover different media or communication technologies as well as the availability of information regarding voter-preferences. We show that it is possible to derive expected values for the development of several variables, like the number of active campaigners and opinions, over time. Going beyond pure modelling we propose fitting the models' parameters to US-data on presidential approval with the aim of deriving a hypothetical "contagiousness" of voter-opinions. This indicator can be used to improve predictions of the development of popularity measures.