Bringing the candidate into models of candidate evaluation

Candidate characteristics are often treated as a lump sum that can be pulled apart but ultimately sum together in models of candidate evaluation. The present study asks, first, whether distinctions among traits are useful for models of candidate evaluation. Second, it considers whether the role of substantive trait dimensions in overall evaluation is uniform across candidates or varies by candidate and electoral context. Latent variable structural equation models are used to test the impact of trait dimensions on thermometer ratings of presidential candidates from 1984 to 1996. Results support the separation of traits into substantive content dimensions. Trait dimensions vary in their effect on candidate evaluations depending on the candidate and election year. This is consistent with a model of attitude response based on changing considerations. These results suggest that as candidates and campaigns call attention to different underlying trait dimensions, the bases for overall evaluation vary accordingly.