Introduction

As discussed in Chapter 1, behavioral researchers and spatial modelers have quite different perspectives on how voters decide. Spatial modelers typically assume that voters are entirely (or at least chiefly) motivated by the policies that the competing parties or candidates present in the current campaign, and, in their empirical applications, policy factors are the chief (if not the only) influences that are incorporated into their models. Behavioral researchers, by contrast, emphasize that voter choices are affected by a variety of considerations besides the parties' current policies, including voters' party identifications and sociodemographic characteristics, their perceptions of economic conditions, and retrospective evaluations of incumbent performance.

Furthermore, while spatial modelers typically posit that voters employ a proximity metric to evaluate parties' policies – that is, that voters prefer parties whose positions are close to their own positions along salient policy dimensions – some research, both theoretical and empirical (see Grofman 1985; Fiorina 1994, 1996; Lacy and Paolino 1998, 2001; Merrill and Grofman 1999; Kedar 2002; Adams, Bishin, and Dow 2004; Lewis- Beck and Nadeau 2004), suggests that this behavior is modified by voters' realization that parties/candidates will probably not be able to implement the full extent of the policies that they advocate. Hence, a proximity model with discounting may be appropriate. As we argue beginning in Chapter 3, both of these perspectives contribute to our understanding of party policy strategies and the nature of mass–elite policy linkages.

In this chapter, we compare the spatial-voting model to the behaviorist's perspective on voting, developing the various components of the unified multivariate voting specification that we use in subsequent chapters to analyze elections in the four countries we study.