Computational Social Science

People care about different things in politics, and when they are asked a question, each individual may process it differently or call on different memories and ideas when formulating their response. In the political science literature, this heterogeneity is usually described as salience: memories, ideas, considerations, or choice features that an individual is likely to apply in his or her decision are considered more salient.

In the universe of spatial ideal point models of ideology, salience takes on a more specific meaning: the weights that people apply to various preference dimensions when making their choices. To the extent that memories and ideas drive people's positions on these dimensions and the likelihood that they will use a particular dimension when making a political choice, the spatial model definition meshes reasonably nicely with the more general definition.

Formal models of voter choice and party strategy capture salience is a very compact and high-level way. Individuals have ideal points in a policy space, they are faced with possible outcomes in the policy space, and their utility of each outcome depends on the salience-weighted distance between their ideal point and the policy of interest. See Hinich and Munger (1997) for a more extensive discussion. Unfortunately, the existing techniques for spatial ideal point estimation are poorly suited to the analysis of survey data, and existing techniques for modeling salience do not operate in the same latent space as salience in formal models.

In applied research on voter choices, salience is almost always related to very specific ideas, considerations, or features of a model. For example, Transue (2007) emphasizes racial identity and examines how that affects issue preferences. Gerber et al. (2010) emphasizes party identity in a field experiment and examines how that influences attitudes. In the related literature on latent moral preferences, Rosenblatt et al. (1989) demonstrate that mortality salience affects how people punish others who violate norms.

These applications are extremely specific and do not relate directly to highlevel summaries of political ideology. The disconnect is apparent in the conclusion of Transue (2007),where the author attempts to connect the policy-specific findings to more general psychological and political theories.