This chapter develops a general stochastic model of elections in which the electoral response is affected by the valence (or quality) of the candidates. In an attempt to explain non-convergence of candidate positions in the 2000 and 2004 Presidential elections, a formal spatial stochastic model, based on intrinsic valence, is presented. A pure spatial model of the election is constructed. It is shown that the equilibria, under vote maximization, do indeed lie at the electoral origin. Other work on Presidential elections in the United States has suggested that a superior empirical model should incorporate the electoral perceptions of the candidate character traits. The chapter then considers a joint model with sociodemographic valences as well as electoral perception of traits and shows by simulation that the vote maximizing equilibrium positions were close to, but not precisely at, the electoral origin. This model used electoral estimates of the candidates.positions. These differed substantially from the estimated equilibria of the traits model. To account for this difference, a more general formal model is then considered where the valence differences between the candidates were due to resources that were contributed to the candidates by party activists. The trade off between activist and electoral support is given by a (first order) balance condition involving, called the centrifugal marginal activist pull. Survey information on party activists, who contributed resources to the candidates, was obtained. It is argued that the difference between the equilibrium obtained from the spatial model with traits, and the estimated candidate positions, is compatible with the location of these activists. The final model is one where the activist resources are used by candidates to target individual voters or groups of voters. The balance condition in this case involves a complex constrained optimization problem, that captures the essence of modern electoral politics.