This paper extends the spatial theory of voting to an institutional structure in which policy choices are a function of the composition of the legislature and of the executive. In an institutional setup in which the policy outcome depends upon relative plurality, each voter has incentives to be strategic since the outcome depends upon how everybody else votes. By applying to this game between voters the refinements of Strong Nash and Coalition Proof Nash we prove existence of equilibria with properties which appear intuitive and realistic. In fact, the model has several testable implications which seem consistent with some observed patterns of voting behavior in the United States and perhaps in other democracies in which the executive is directly elected. For instance, the model predicts: a) split-ticket voting; b) for some parameter values, a split government with different parties controlling the executive and the majority of the legislature; and c) the mid-term electoral cycle.