Knowledge and Political Categorization

Knowledge and Political Categorization Evan Heit (eheit@ucmerced.edu) Stephen P. Nicholson (snicholson@ucmerced.edu) School of Social Sciences, Humanities and Arts University of California, Merced Merced CA 95343 USA Abstract A nationally representative sample of US adults completed two political categorization tasks. The first was to identify the political parties for hypothetical candidates with information given about demographics and stands on issues. The second task was to decide whether to vote for each candidate. On the identification task, judgments about whether a person is a Democrat were almost a perfect mirror image of judgments of whether a person is a Republican. In general, respondents were very successful in the identification task; there was a strong correlation with objective probabilities. Likewise, respondents were successful at the voting task, in terms of their own party interests. Success at these two tasks was positively correlated with a measure of political knowledge. The pattern of responses was also influenced by the political party of the respondent; suggesting that feature weights depended on party membership. Implications for models of categorization and reasoning are discussed. Keywords: Categorization; Expertise; Probability Judgment; Political Cognition. Introduction We propose that political parties should be conceived of as categories. Following Rosch & Mervis’s (1975) seminal work on categorization, political parties have a horizontal dimension corresponding to typicality structure, e.g., Mitt Romney is a more typical Republican than is Ron Paul. It is then appropriate to ask what is the function of political categories (cf., Anderson, 1991; Billman & Heit, 1988; Markman & Ross, 2003), beyond labeling individuals as party members. One key function is to support voting, which can be seen as a category-based inference, e.g., knowing that Mitt Romney is a typical Republican would lead many people to vote for him in a Presidential election. In previous research (Heit & Nicholson, 2010) we have collected typicality judgments for a set of real political candidates. College students rated the individuals either on typicality as a Democrat or typicality as a Republican. The relation between the two sets of ratings was strong, negative, and linear, with a remarkable correlation of -0.9957. Essentially, whatever made an individual more typical of one party was seen to make that individual less typical of the other party (cf., Rosch & Mervis, 1975; Verbeemen, Vanoverberghe, Storms, and Ruts, 2001). It was not possible to be typical of both parties, or atypical of both parties. The results contrasted with other opposing pairs of categories, male versus female jobs and healthy foods versus junk foods. We concluded that for political categories, there is a highly systematic and polarized representation of knowledge. Although the results were extremely strong, the study itself had limitations. For example, students may not be representative of voters at large. We did not systematically study the effects of demographic variables such as level of political knowledge (which might be low for college students) and party of the respondent. Because the stimuli were simply names of public figures, we could not tell which information about these figures was being used. Also, the dependent variable, typicality, has disadvantages, because it is not objective and it may not map directly onto real political behavior such as voting. Hence, the present experiments substantially improved upon Heit and Nicholson (2010). Each experiment involved several hundred adults from a nationally representative sample of US adults, with information collected about political knowledge and party membership. The stimuli were descriptions of hypothetical candidates in terms of demographic information (race, gender, number of children) and stands on issues (government spending and abortion). Information about each candidate’s political party was omitted from the stimuli; however the objective probability of being a Democrat or Republican based on demographics and stands on issues could be determined from national survey data. In Experiment 1, the task was to identify each candidate’s party. In effect, we were examining whether respondents could correctly categorize candidates as Democrats or Republicans when this information is withheld. In Experiment 2, the task was voting; respondents were asked how likely they would be to vote for each candidate. A key measure of interest was whether respondents voted the party line, i.e., Democrats voting for Democrats and Republicans voting for Republicans. In general, we were interested in whether performance on these two tasks depended on political knowledge and party membership of the respondent. We also examined the influence of various cues, to see if different cues were used for the two tasks and by different sub-groups of respondents. In the cognitive science literature on categorization, perhaps the most closely related work addresses the effects of expertise on biological categorization. For example, Johnson and Mervis (1997) studied categorization of