Testing the Proximity and Directional Models of Voter Utility within a Nested Statistical Framework

Theoretically, we have shown that proximity, directional, and intensity components can all be embedded in a two-parameter unified model of the shape of voter utility functions, of which the traditional proximity spatial model, the Rabinowitz–Macdonald (RM) model, and the Matthews directional model are all special cases. In developing this unified model we showed that the RM directional model, despite its name, is best viewed as a mixed model with both directional and intensity components, whereas the Matthews model best represents a pure directional component.

To provide a preliminary test of this unified model, we look at the nature of voter utility functions for major candidates for the U.S. presidency during the period l980–96 by fitting the unified model developed in Chapter 3 to American NES data. In Chapters 6 and 7 we develop and test a unified model of voter choice for these and other data.

Not unexpectedly, by testing nested submodels, we conclude that, in most applications, pure models must be rejected in favor of a model incorporating two or more of the proximity, directional, intensity, or discounting components. Also – like Enelow, Endersby, and Munger (l993) – we provide evidence that different models of utility functions best explain voter utility for different types of candidates.