From Stokes's (1963) early critique on, it has been clear to empirical researchers that the traditional spatial theory of elections is seriously flawed. Yet fully a quarter century later, that theory remains the dominant paradigm for understanding mass-elite linkage in politics. We present an alternative spatial theory of elections that we argue has greater empirical verisimilitude. Based on the ideas of symbolic politics, the directional theory assumes that most people have a diffuse preference for a certain direction of policy-making and that people vary in the intensity with which they hold those preferences. We test the two competing theories at the individual level with National Election Study data and find the directional theory more strongly supported than the traditional spatial theory. We then develop the implications of the directional theory for candidate behavior and assess the predictions in light of evidence from the U.S. Congress.
Rabinowitz et al. (1989) studied this question.