The Dimensionality of Parliamentary Voting

Abstract

The chapter discusses different ways to estimate the dimensionality of roll-call voting data. These methods use data from the U.S. House of Representatives, and the author shows that there were periods when a two-dimensional representation was necessary and others when a one-dimensional representation captures all but a relatively small percentage of the variance. The author then considers data from the UN General Assembly from before the fall of the Berlin Wall, finding a communist vs. anti-communist dimension and a pro- and anti-Israel dimension, as well as data from the French National Assembly early in the 5th Republic that finds a one-dimensional representation fits nearly perfectly. The author then considers some more technical issues about best methods, concluding that there is no foolproof way of determining the “true” dimensionality of a roll-call matrix, and no substitute for substantive understanding of the politics and policy shaping the roll calls.