Evolution of beliefs in a society is a result of interactions between people in the society over generations. We analyze the long term dynamics of belief evolution by combining people’s prior beliefs, social dynamic network structures and the confusion that occurs between beliefs. The main contribution of this work is threefold. First, we explore the belief evolution using existing network models such as scale free networks and small world networks to create social communication structures and belief confusion structures. Second, we model the belief evolution with homophily based models using different statistical distances. We compare the individual and societal belief distributions and trends obtained from those models. Third, we explore the evolution of religious affiliations in different countries; both large and small in size, located in different continents. We use a homophily based model to fit religious affiliation data to model the dynamics of religious beliefs in Australia, Canada and Ireland over time. Using existing network models, we observe that the society evolves to a homogeneous belief system. However, the formation of heterogeneous belief systems such as social groups that share the same beliefs and isolated individuals can be observed using homophily based models. Moreover we can see how those formations change based on different distance measures. Finally, the successful implementation of real world data justifies the theoretical formulation of the model. This allows us to interpret the social dynamics over time as well as their future implications.