This paper describes two functions for simulating voter decisions across two elections, based on the overdispersed Multinomial model by Forcina and colleagues [8]—a model that enables the generation of individual-level electoral data under the assumption of known parameters, particularly transition probabilities. Compared to other voter simulation approaches, this model offers a more realistic depiction of voter behavior by incorporating the assumption that, within each local unit, individuals who supported the same option in the first election tend to cluster into social circles of relatives, friends, or neighbors, whose members make correlated decisions through everyday interactions. The first function assumes homogeneity (no covariates) in transition probabilities across units. The second function introduces heterogeneity in transition probabilities, simulating individual choices under the assumption of a varying presence across local units of strongly loyal, (trend-following) strategic, and context-influenced voters. The simulated data include the joint frequency distribution of voters across the two elections—an outcome impossible to observe in reality due to the secrecy of voting. These functions are implemented in the R package eiCircles, facilitating their use in research and practical applications in electoral modeling.