AGGREGATE DATA IN ELECTION PREDICTION

In: Public opinion quarterly: journal of the American Association for Public Opinion Research, Band 33, Heft 1, S. 96-99

Abstract

The election night vote estimation device used in the 1962 Iowa gubernatorial election by Robert W. Clyde, William J. Hemmerle, & T. A. Bancroft (see SA 0827/A8902) was modified & tested by the author in the 1966 Minn gubernatorial election. This system represents a low-cost method for estimating rapidly & accurately final vote ratios in statewide election contests. Minn's 87 counties were placed into homogeneous groupings on the basis of pop & historical voting patterns. Pop statistics from the 1960 census were used. Each county's total Democratic vote was added up for the past 3 gubernatorial elections & divided by the total votes cast, thus yielding a 'historical %' for each county. Strata were defined by pop & historical % to create a minimum N of cells necessary to yield sufficient representation early in the evening, yet enough cells to satisfy reasonably the assumption of homogeneity. The values of estimated final Democratic % & estimated final Republican % were established by computer. County-by-county returns were telephoned to the authors at the computer site throughout election night. Results lead to the conclusion that it is possible to predict aggregate electoral behavior with a high degree of accuracy without explicit reliance on SE variables. No such factors as age, religion, income, occup, etc, were included. It is suggested that in rapid vote estimating, broader sampling categories (eg, historical tendencies) may minimize some of the disruptive & transitory variables which often haunt samples based upon narrower & more select SE factors. 2 Tables. M. Maxfield.