A New Approach to Pre-Election Polling

Abstract

PRE-ELECTION POLLS currently utilize either a simulation approach or an attitude approach. Polls based on the simulation approach seek to determine the distribution of potential votes among candidates or parties at various points in time prior to election day. This type of poll is used mostly by professional pollsters (Mendelsohn and Crespi, 1970). On the other hand, polls based on the attitude approach try to ascertain the degree of support among voters for various candidates and parties. They are employed mainly by political scientists to study the voter dynamics and to attempt to predict voter behavior (Campbell et al., 1960; Kelley and Mirer, 1974; Dobson and St. Angelo, 1975). Polls based on the simulation and attitude approaches can make use of two dis-tinct types of measurement: voters' intentions and voters' attitudes. Clearly, the simulation approach relies heavily on voters' intentions, with attitude measurement, if it appears at all, providing complementary information (Teer and Spence, 1973: 210). The opposite is true for polls based on the attitude approach (Kelley and Mirer, 1974:576). Unfortunately, it is not possible to obtain a perfect correlation for the two types of measurement, since they are derived independently and utilize different types of scales. At best the transfer function from one scale to another is probabilistic on a post-facto bais (Kelley and Mirer, 1974:582). No single measure, however, yields both types of results simultaneously.