Multiple Ideal Points: Revealed Preferences in Different Domains

Abstract We extend classical ideal point estimation to allow voters have different preferences when voting in domains—for example, on agricultural policy than defense policy. Our scaling procedure results estimated points a common scale. As result, we are able directly compare member’s revealed across domains of (different sets motions) assess if, for member votes more conservatively agriculture motions defense. In doing so, the extent which behavior an individual voter is consistent with uni-dimensional spatial model—if has same all domains. The key novelty estimate rather assume identity “stayers”—voters whose preference constant votes. approach offers methodology investigating relationship between basic space and issue legislative (Poole 2007). There several methodological advantages our approach. First, model allows testing sharp hypotheses. Second, developed can be understood as kind partial-pooling item response theory scaling, resulting less uncertainty estimates. Related, method provides principled unified “granularity” (i.e., level aggregation) analysis roll-call data (Crespin Rohde 2010; Roberts et al. 2016). illustrate by estimating U.S. House Representatives members’ domains, identify other potential applications including: studying committee floor behavior; constituency influence representation.