A re-evaluation of alternate methodologies in legislative voting analysis

Abstract

This paper consists of a comparison of Guttman scale analysis and factor analysis as applied to legislative roll-call votes. After defining each technique, their respective shortcomings are discussed. Guttman scaling has intuitive appeal, but lacks the precision of measurement and ability to produce interval level legislator scales that factor analysis provides. Factor analysis is frequently criticized, however, because it cannot generate factors that resemble Guttman scales in composition. By reconsidering the basis for each technique, two points apparently overlooked by researchers are established. Firstly, although the techniques produce different results, both kinds are meaningful in their own right. Secondly, the techniques are not just different; they are interrelated in a definable manner. This interrelationship is developed by means of a principal components analysis of artificial roll-calls forming a perfect Guttman scale. The results are then used to establish probable interpretations of the principal components of a scalable set of roll-calls. The implication is that the precision of factor analysis need not be sacrificed even if the scaling criterion is preferred by the analyst.