Closeness Counts Only in Horseshoes and Dancing

During the period in which our article (APSR vol. 68 [June 1974]) circulated in manuscript form it provoked an unusual amount of collegial reaction. Of course, we were quite prepared for a reaction from those who use decision-theoretic models in their research—they were our intended audience. More surprisingly, we also received comments from less directly involved bystanders—a medieval historian for example. All this correspondence indicates to us that nearly everyone has his own theory of how voters behave, and that most such theories do not agree with the one presented in our article. The comments of Professors Tullock, Beck, Mayer and Good, and Stephens further support this conclusion. In an appendix to this note we have responded to the imaginative point raised by Tullock. As for the traditional questions raised by our other critics, however, we adopt a different line of rebuttal. Rather than conduct an unfruitful debate over the a priori plausibility of the minimax regret model we will do something that theorists too seldom do: examine some data. Before doing so we will make an important distinction between using a model prescriptively and using it descriptively. (Decision-theoretic types tend to move a bit too easily from one usage to the other.) Then, after reviewing the major point of our article we will turn to the data.