Elections embody valuable information on the dynamics through which individuals influence each other and make choices. We analyze proportional elections in Brazil, a country with a diversified and huge electorate of around 100 million people, covering the period 1970-2014. Through the distribution p(v) of the number of candidates receiving v votes, we perform a comparative analysis of different elections in the same calendar and as a function of time. Inspired in multi-species population dynamics, we propose a model, consisting in a system of nonlinear differential equations with stochastic parameters, that allows to understand and interpret the observed features. We show that the distribution of votes among candidates reveals patterns that reflect the evolution of people interactions. In particular, a statistical property of vote distributions, that appears to be predominantly associated to the electorate, can be used as a measure of its degree of feed-back. Such measure of a population, which is hard to be accessed otherwise, may be useful to know the extent to which people interact and can be influenced, even beyond the context of political elections.