Abstract
Citizens are asked to make many judgments in politics, often in the face of scarce information and limited motivation. In making political judgments, citizens may rely upon a variety of cues, including the partisanship, ethnicity, race, or sex of candidates. Some cues, however, are more democratically troublesome than others. Democratic norms of equality suggest that attitudes towards racial or ethnic groups should not influence citizens’ evaluations of candidates. Often, however, attitudes towards these groups do matter. This article identifies a limiting condition on the effect of group attitudes: the presence of a party cue. I demonstrate that attitudes towards Hispanics influence willingness to support a Hispanic candidate, but only in the absence of a party cue. The article also contributes to existing work by analyzing both explicit and implicit measures of attitudes towards groups. Explicit measures include stereotypes and feeling thermometers; implicit measures are derived from a subliminal priming task. Subjects with positive attitudes towards Hispanics (whether these attitudes were measured implicitly or explicitly) were more likely to support the Hispanic candidate, in the absence of party cues. Subjects with negative attitudes towards Hispanics were less likely to support the Hispanic candidate, in the absence of party cues. The presence of party cues, however, eliminates the impact of attitudes towards Hispanics on political choice.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
For example, Fazio et al. (1995) suggest that whites’ implicit attitudes towards blacks are correlated with less friendly interactions between white subjects and a black experimenter. Wilson, Lindsey, and Schooler (2000) find that implicit measures of hostility towards blacks are correlated with less physical contact between a white subject and a black confederate.
Citizens may also support candidates as a result of perceptions of shared demographic characteristics with the candidate. As Sigelman and Sigelman (1984) suggest, “The literature on interpersonal attraction... provides abundant evidence that people consistently use similarity to themselves as a major basis for finding others attractive or unattractive” (264). One mechanism underlying this proposition could be that voters who share similar demographic characteristics with a candidate believe that what the candidate does in office will benefit them; conversely, voters who do not share those characteristics may see such a candidate as posing a threat to them. The present sample’s composition makes it difficult to test this proposition with much certainty; however, the data analysis suggests that Hispanic ethnicity predicts willingness to support a Hispanic candidate, when group attitudes are not included in the model. The ethnicity effect fades away, however, once attitudes towards Hispanics as a group are incorporated into the model. Hence, a group-attitudes account is more consistent with the data than a shared demographics account.
Voters might also use perceptions of candidates’ demographic characteristics as a way of predicting what policy actions the candidate will take in office; as Pitkin (1967) notes, “We tend to assume that people’s characteristics are a guide to the actions they will take” (89). Similarly, McDermott (1997) notes, “If voters stereotype candidates by group affiliation, then demographic characteristics should provide information based on commonly held social stereotypes” (271). The candidate’s race/ethnicity might thus activate group-based stereotypes which then shape the voter’s inferences about the candidate’s ideological positions, issue preferences, traits, or competence in handling issues.
There has been little research on voters’ reactions to candidates of other ethnic backgrounds. One exception is Sigelman, Sigelman, Walkosz, and Nitz (1995), which includes one Hispanic candidate in their experimental manipulations.
Subjects were recruited in two ways: through invitation letters sent to a representative group of 1,000 individuals living in nearby towns; and through an email invitation sent to a randomly selected group of 650 non-faculty, non-research university staff. Subjects were representative of the local area. 76% of subjects were white (10% Asian and 8% Hispanic). The modal age was between 46–55 years old (27% of the sample). 63% of subjects were female. 71% of the sample had a bachelor’s degree or higher, which is only 3% higher than the Census 2000 data on the proportion of residents holding bachelor’s degrees or higher. 31% of subjects identified as Republicans (strong, weak, or leaning) and 45% identified as Democrats (strong, weak, or leaning).
Each subject was seated at his/her own computer terminal. The experimental laboratory hosts six computer terminals. Most sessions included about four subjects. Subjects were scheduled for 45 minute time slots, but most completed the study in about 30 minutes.
These questions, along with a political information battery, were included to disguise the true nature of the study. They are not analyzed in this article.
The election of judges in other states is partisan, thus allowing for applicability of the present design. These states are: Alabama, Illinois, Louisiana, Michigan, Ohio, Pennsylvania, Texas, and West Virginia (American Judicature Society, 2004). Further, a 2002 ruling by the United States Supreme Court struck down a Minnesota law that prohibited judicial candidates from taking stands on issues and affiliating themselves with political parties (Republican Party of Minnesota v. White), thus opening the door for judicial candidates to ally themselves with particular stands or parties.
A choice has to be made regarding whether the stimulus materials focus on fictional or real candidates. With fictional candidates, researchers can minimize prior information and maximize control over the stimulus. However, there is an obvious lack of realism. Real candidates are troublesome in that people bring real (and differing amounts of) information into the experiment, and they constrain the reasonable set of manipulations that can be imposed on subjects. I elected to use real candidates, in order to bolster the external validity of the study. The tradeoff, however, is that this is not a fully factorial design: the direction of partisanship is not manipulated; only the presence or absence of party cues is manipulated. I return to this issue in the concluding section of the article.
The order in which the candidates appeared (from left to right) was randomized, in order to minimize order effects.
To be precise, the candidate information did not explicitly note that Kathryn Werdegar and Marvin Baxter were White. It is possible that subjects could have ascribed to these candidates another race or ethnicity. I proceed with the assumption that most subjects would indeed have presumed both to be White, given the availability heuristic concerning judges. Pilot testing suggests that the name “Marvin” is not strongly associated with Blacks as a group.
Ten subjects (4.3% of the sample) elected the “No Response” option. Consistent with past research that shows that source cues assist in opinion-holding (Mondak, 1993; Squire & Smith, 1988), these subjects were disproportionately found in the no party cues condition (eight of the ten subjects). They are omitted from analysis.
In order to minimize nonrandom response error attributable to measurement, the direction of the traits was reversed for trustworthy- not trustworthy. Subjects were asked to rate eight groups (Whites, Blacks, Hispanics, Asians, women, men, Democrats, and Republicans), in randomized order. This battery has appeared in the General Social Survey and the National Election Studies. See Smith (1990) for additional details.
The additive scale could range from −1 to 1. In practice, the stereotype-based attitudes towards Hispanics range from −0.67 (negative) to 1 (positive), with a mean of 0.20, standard deviation of 0.30, and α = 0.53. Subjects self-reporting as Hispanic register significantly more positive attitudes towards Hispanics compared with non-Hispanic subjects (p < 0.006, one-tailed).
The stereotype battery appeared after the subliminal priming measures were administered. No procedures were put in place to prevent the priming task from influencing the explicit measures. However, a multitude of groups were incorporated into the subliminal priming measures, and these groups were paired with a balance of positive and negative targets. Additionally, a cognitively taxing task (a political information battery) was included as a distractor task, sandwiched between the stereotype battery and the subliminal priming task. Hence, it is not clear whether the stereotype battery could have been contaminated by the responses in the subliminal priming task. To be more precise, whether the contamination would skew the responses systematically in one way or another is hard to see.
See https://implicit.harvard.edu/implicit.
Note that several political science studies analyze the consequences of implicit activation of stereotypes: e.g., Berinsky and Mendelberg (2005); Mendelberg (2001). In those studies, the activation of stereotypes occurs through experimental manipulation, and the activation, or treatment, is used as an independent variable. Here, the intention is to use subliminal priming to measure attitudes towards groups and to use those attitudes as independent measures.
The exact instructions were as follows: “Just before each word that you are to categorize you will see one or more words and letter strings briefly flashed. It is your task to IGNORE these brief flashes. Respond only to the last, clearly visible word shown on the screen. Press the space bar to continue. [next screen] When you press the space bar, you will see a word to which you should respond. As a reminder of the instructions for responding: Press the 'f' key if you see a PLEASANT word. Press the 'j' key if you see an UNPLEASANT word. Please answer the questions as quickly as you are able. If you are unable or unwilling to record an evaluation, just press the space bar. Press the space bar to continue.”
The entire experiment (including survey questions and the subliminal priming task) was administered using Inquisit 2.0, a psychological software package designed to allow for precise response latency measurement. See http://www.millisecond.com for additional details.
Eight groups were used in the experiment, in randomized order. The eight groups were: Whites, Blacks, Hispanics, Asians, women, men, Democrats, and Republicans. The groups appeared in capital letters, in the center of the screen.
The list of 399 normed words is available at: http://faculty.washington.edu/agg/pdf/bgb.txt. The words are rated in terms of pleasantness (from pleasant to unpleasant). The decision rules for selecting the target words were as follows: Start with the 25 most pleasant words and 25 least pleasant words. Discard words with more than two syllables or more than eight letters. Discard words that are stereotypically (or otherwise) associated with groups in the study. Randomly select words for neutral practice. Practice words were: sunrise, heaven, honest, sunset, vomit, lice, poison, sickness. Target words in the 32 trials that were paired with the group primes were: love, laughter, kiss, pleasure, joy, life, happy, peace, rainbow, freedom, friend, hug, truth, cheer, angel, beach, tumor, torture, smallpox, rabies, devil, slaughter, maggot, cancer, scum, ulcer, death, hatred, failure, dreadful, prison.
More trials would have yielded less noisy measures at the individual-level. With fewer trials, the test is more conservative and is thus less likely to produce significant results. That significant results are found (as reported, later) with so few trials should make the results even more compelling. The large sample size also helps to ameliorate this random noise at the individual-level.
As is advised with latency data, the data were cleaned, following the protocol discussed in Lodge et al. (2003): (1) Eliminate all responses where reaction time <300 ms; (2) Recode all responses where reaction time >2500 ms to 2500 ms; and (3) Eliminate all incorrect responses (space bar or incorrect valence).
Hispanics score much higher on this measure, with a mean of 287.0 and standard error of 106.0, whereas the mean among non-Hispanics is 142.0, with a standard error of 36.3. Small sample size prevents this difference from achieving statistical significance at conventional levels (there are only 17 Hispanics compared with 215 non-Hispanics); the p-value for the difference of means test is a one-tailed p < 0.11.
Karpinski and Hilton (2001) argue that the IAT captures “environmental associations”—that is, cultural associations between groups and concepts—rather than personally endorsed beliefs. Karpinski and Hilton also note that this argument cannot necessarily be generalized to attitudes measured using a subliminal priming procedure (787).
I return to this discussion of the relationship between attitudes measured using implicit and explicit instrumentation in the Conclusion.
All variables are scaled zero to one with the exception of the implicit measure. In order to preserve the meaning of zero in the implicit measure, it was divided by the absolute value of the maximum value. This imposes an upward bound of + 1 on the measure and establishes 0 as the “neutral” point. Hispanic is a dummy: 1 if Hispanic; 0 else. Female is a dummy: 1 if female; 0 if male. Party identification is trichotomous: 0 if Republican; 0.5 if Independent; 1 if Democrat. Analyses with 7-category measure of party identification were similar. Party cue is a dummy: 1 if party cue condition; 0 if baseline.
Predicted probabilities are generated holding female and Hispanic to zero. The 90% confidence intervals for these predicted probabilities are [0.487,0.916] and [0.096, 0.524], respectively. I utilize 90% confidence intervals given the relatively small sample size in this study.
The 90% confidence interval around the predicted probability of 0.267 is [0.096,0.524]; the 90% confidence interval around the predicted probability of 0.829 is [0.591,0.953].
The 90% confidence intervals for these predicted probabilities are [0.057, 0.424] and [0.610, 0.973] respectively. Interpretation of the effects of the group-based attitudes (both explicit and implicit) provides some suggestion of overcorrection effects. The effect of group-based attitudes becomes attenuated and even appears to a limited extent to slide in the direction of overcorrection (however, the effects of overshooting are not statistically distinguishable from zero, so the results are merely suggestive in this regard). This phenomenon is consistent with the literature on assimilation versus contrast effects. These various models include the “set-reset” model proposed by Martin and colleagues (Martin, 1986; Martin, Seta, & Crelia, 1990); the “affect as information” and inclusion-exclusion model proposed by Schwarz and colleagues (Schwarz & Bless, 1992; Schwarz, Strack, & Mai, 1991); the misattribution and correction model proposed by Ottati and Isbell (1996); and the Flexible Correction Model proposed by Wegener and Petty (Wegener & Petty, 1995; Petty, Wegener, & White, 1998). These models, though they differ in the details, suggest that some individuals, under some circumstances, may be motivated to suppress or overcompensate for the possible influence of a particular prime or piece of information on their judgments. Applied to the present case, the presence of the party cue might have driven subjects to overcorrect for the possible influence of attitudes towards Hispanics on their choice of candidate.
Respondents were asked about a series of eight groups, including Hispanics, administered in randomized order. The mean feeling thermometer rating for Hispanics was 75.02, with a standard deviation of 21.31. The feeling thermometer rating correlates positively with the stereotype trait ratings: at 0.41, but, like the stereotype trait rating, barely correlates with the implicit measure (r = −0.03). Subjects who self-identify as Hispanics report significantly higher ratings of Hispanics compared with non-Hispanic subjects (a mean of 85.6 and standard error of 3.54, versus a mean of 74.1 and standard error of 1.49, difference of means statistically significant at p < 0.01, two-tailed). In the probit analysis, this measure is rescaled to range from −1 (0 degrees) to 1 (100 degrees), with a score of 0 indicating a neutral 50 degrees.
The same clarity and consistency of effects does not emerge with respect to attitudes (explicit or implicit) towards other groups (women, as shown in Table 5; attitudes towards Democrats and Republicans, analyzed separately, and available upon request).
For a related, more recent line of work on justification and suppression of prejudice, see Crandall and Eshleman (2003).
A fully-factorial design with three candidates, where sex of the candidates, ethnicity of the candidates, the partisanship of the candidates (along with the presence or absence of party cues) would require 63 conditions: Assuming three candidates, we would need a 3 (3 possible combinations where one candidate is female and the other two are male) × 3 (3 possible combinations where one candidate is Hispanic and the other two are White) × 7 (6 possible combinations of partisanship, assuming no single-party slates, plus one No Party Cue condition) design. This number far exceeds what is feasible for most laboratory experiments, so future designs might still run into limitations in adjudicating between different mechanisms.
References
American Judicature Society (2004). Judicial selection in the states: Appellate and general jurisdiction courts. Des Moine, IA: American Judicature Society. http://www.ajs.org/js/JudicialSelectionCharts_old.pdf. Last accessed 8/19/04. .
Banaji, M. R., & Greenwald, A. G. (1994). Implicit stereotyping and prejudice. In Mark P. Zanna, & James M. Olson (Eds.), The psychology of prejudice (pp. 55–76). Hillsdale, NJ: Lawrence Erlbaum Associates.
Bellezza, F. S., Greenwald, A. G., & Banaji, M. R. (1986). Words high and low in pleasantness as rated by male and female college students. Behavioral Research Methods, Instruments, and Computers, 18, 299–303.
Berinsky, A. J. (2004). Can we talk? Self-presentation and the survey response. Political Psychology, 25, 643–659.
Berinsky, A. J., & Mendelberg, T. (2005). The indirect effects of discredited stereotypes in judgments of Jewish leaders. American Journal of Political Science, 49, 845–864.
Bruner, J. S. (1957). On perceptual readiness. Psychological Review, 64, 123–52.
Burdein, I., Lodge, M., & Taber, C. (2006). Experiments on the automaticity of political beliefs and attitudes. Political Psychology, 27, 359–371.
California Secretary of State (2002). California general election, Tuesday, November 5, 2002, Official voter information guide. http://vote2002.ss.ca.gov/2002-vig/pdf/bp_pe01.pdf. Last accessed 8/19/04. .
Carlston, D. E., & Smith, E. R. (1996). Principles of mental representation. In E. T. Higgins, & A. W. Kruglanski (Eds.), Social psychology: Handbook of basic principles (pp. 184–210). New York: Guilford.
Citrin, J., Green, D. P., & Sears, D. O. (1990). White reactions to black candidates: When does race matter? Public opinion quarterly, 54, 74–96.
Crandall, C. S., & Eshleman, A. (2003). A justification-suppression model of the expression and experience of prejudice . Psychological Bulletin, 129, 414–446.
Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and controlled components. Journal of Personality and Social Psychology, 56, 5–18.
Dovidio, J. F., Kawakami, K., Johnson, C., Johnson, B. et al. (1997). On the nature of prejudice: automatic and controlled processes. Journal of Experimental Social Psychology, 33, 510–540.
Draine, S. C., & Greenwald, A. G. (1998). Replicable unconscious semantic priming. Journal of Experimental Psychology, 127, 286–303.
Dubois, P. L. (1984). Voting cues in nonpartisan trial court elections: A multivariate assessment. Law and society review, 18, 395–436.
Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. (1995). Variability in automatic activation as an unobtrusive measure of racial attitudes: a bona fide pipeline?. Journal of Personality and Social Psychology, 68, 1013–1027.
Fazio, R. H., & Olson, M. A. (2003). Implicit measures in social cognition research: Their meaning and use. Annual Reviews of Psychology, 54, 297–327.
Fiedler, K., & Bluemke, M. (2005). Faking the IAT: Aided and unaided response control on the Implicit Association Tests. Basic and Applied Social Psychology, 27, 307–316.
Gaertner, S. L., & Dovidio, J. F. (1986). The aversive form of racism. In John F. Dovidio, & Samuel L. Gaertner (Eds.), Prejudice, discrimination, and racism (pp. 61–89). Orlando, FL: Academic press.
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 74, 1464–1480.
Kam, C. D. (2005). Who toes the party line? Cues, values, and individual differences. Political Behavior, 27, 163–182.
Karpinski, A., & Hilton, J. L. (2001). Attitudes and the implicit association test. Journal of Personality and Social Psychology, 81, 774–788.
Karpinski, A., & Steinman, R. B. (2006). The single category implicit association test as a measure of implicit social cognition. Journal of Personality and Social Psychology, 91, 16–32.
Kihlstrom, John F. (2004). Implicit methods in social psychology. In C. Sansone, C. C. Morf, & A. T. Panter (Eds.), Handbook of methods in social psychology (pp. 195–212). Thousand Oaks, CA: Sage.
Lodge, M., & Taber, C. (2000). Three steps toward a theory of motivated political reasoning. In A. Lupia, M. D. McCubbins, & S. L. Popkin (Eds.), Elements of reason (pp. 183–213). Cambridge: Cambridge University Press.
Lodge, M., Taber, C., Burdein, I., & Yener, S. (2003). The impact of self-identifications on political attitudes: An experimental test employing subliminal priming. Presented at the Annual Meeting of the American Political Science Association, Philadelphia.
Lodge, M., & Taber, C. S. (2005). The automaticity of affect for political leaders, groups, and issues: An experimental test of the hot cognition hypothesis. Political Psychology, 26, 455–482.
Maheswaran, D., & Chaiken, S. (1991). Promoting systematic processing in low-motivation settings: Effect of incongruent information on processing and judgment. Journal of Personality and Social Psychology, 61, 13–25.
Martin, L. L. (1986). Set/reset: Use and disuse of concepts in impression formation. Journal of Personality and Social Psychology, 51, 493–504.
Martin, L. L., Seta, J. J., & Crelia, R. A. (1990). Assimilation and contrast as a function of people’s willingness and ability to expend effort in forming an impression. Journal of Personality and Social Psychology, 59, 27–37.
McDermott, M. L. (1997). Voting cues in low-information elections: Candidate gender as a social information variable in contemporary United States elections. American Journal of Political Science, 41, 270–283.
McDermott, M. L. (1998). Race and gender cues in low-information elections. Political Research Quarterly, 51, 895–918.
Mendelberg, T. (2001). The race card. Princeton: Princeton University Press.
Mondak, J. J. (1993). Public opinion and heuristic processing of source cues. Political Behavior, 15, 167–192.
Nisbett, R., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental processes. Psychological Review, 84, 231–258.
Nosek, B. A. (2005). Moderators of the relationship between implicit and explicit attitudes. Journal of Experimental Social Psychology, 134, 565–584.
Nosek, B. A., & Banaji, M. R. (2001). The go/no-go association task. Social Cognition, 19, 625–664.
Nosek, B. A., & Banaji, M. R. (2003). (At least) two factors moderate the relationship between implicit and explicit attitudes. In R. K. Ohme, & M. Jarymowicz (Eds.), Natura Automatyzmow (pp. 49–56). Warszawa: WIP PAN & SWPS.
Ottati, V., & Isbell, L. M. (1996). Effects of mood during exposure to target information on subsequently reported judgments: An on-line model of misattribution and correction. Journal of Personality and Social Psychology, 71, 39–53.
Petty, R. E., Wegener, D. T., & White, P. H. (1998). Flexible correction processes in social judgment: Implications for persuasion. Social Cognition, 16, 93–113.
Pitkin, H. F. (1967). The concept of representation. Berkeley, CA: University of California.
Popkin, S. L. (1994). The reasoning voter (2nd ed.). Chicago: University of Chicago.
Rahn, W. (1993). The role of partisan stereotypes in information processing about political candidates. American Journal of Political Science, 37, 472–496.
Republican Party of Minnesota v. White. (2002). 536 U.S. 765.
Schwarz, N., & Bless, H. (1992). Constructing reality and its alternatives: An inclusion/exclusion model of assimilation and contrast effects in social judgment. In L. L. Martin, & A. Tesser (Eds.), The construction of social judgments (pp. 217–48). Hillsdale, NJ: Lawrence Erlbaum Associates.
Schwarz, N., Strack, F., & Mai, H. (1991). Assimilation and contrast effects in part-whole question sequences: A conversational logical analysis. Public Opinion Quarterly, 55, 3–23.
Sears, D. O. (1993). Symbolic Politics. In S. Iyengar, & W. J. McGuire (Eds.), Explorations in political psychology (pp. 113–49). Durham: Duke University Press.
Sigelman, C. K., Sigelman, L., Walkosz, B. J., & Nitz, M. (1995). Black candidates, white voters: Understanding racial bias in political perceptions. American Journal of Political Science, 39, 243–265.
Sigelman, L., & Sigelman, C. K. (1982). Sexism, racism, and ageism in voting behavior: An experimental analysis. Social Psychology Quarterly, 45, 263–269.
Sigelman, L., & Welch, S. (1984). Race, Gender, and Opinion Toward Black and Female Presidential Candidates. Public Opinion Quarterly 48(2), 467–475.
Smith, T. W. (1990). Ethnic images: GSS Topical Report No. 19. General Social Surveys, National Opinion Research Center, University of Chicago. http://webapp.icpsr.umich.edu/GSS/rnd1998/reports/t-reports/topic19.htm. .
Squire, P., & Smith, E. R. A. N. (1988). The effect of partisan information on voters in nonpartisan elections. Journal of Politics, 50, 169–179.
Steffens, M. C. (2004). Is the implicit association test immune to faking?. Experimental Psychology, 51, 165–179.
Terkildsen, N. (1993). When white voters evaluate black candidates: The processing implications of candidate skin color, prejudice, and self-monitoring. American Journal of Political Science, 37, 1032–1053.
Wegener, D. T., & Petty, R. E. (1995). Flexible correction processes in social judgment: The role of naive theories in corrections for perceived bias. Journal of Personality and Social Psychology, 68, 36–51.
Wilson, T. D., Lindsey, S., & Schooler, T. Y. (2000). A model of dual attitudes. Psychological Review, 107, 101–126.
Acknowledgements
Many thanks to Bob Huckfeldt, Don Kinder, Tom Nelson, and seminar participants at the Social Psychology/Personality Brown Bag series at the University of California, Davis and at the Micro-Politics Group at the University of California, Davis. This work has also benefited from the constructive advice of three anonymous reviewers. I gratefully acknowledge financial support from the Institute for Governmental Affairs at the University of California, Davis. I thank Idin Eftekhari, Josh Maxwell, Carl Palmer, Chris Rheinheimer, and Brandon Storment for research assistance. All errors remain my own.
Author information
Authors and Affiliations
Corresponding author
Additional information
This paper was originally presented at the 2004 Annual Meetings of the American Political Science Association, Chicago, IL.
Rights and permissions
About this article
Cite this article
Kam, C.D. Implicit Attitudes, Explicit Choices: When Subliminal Priming Predicts Candidate Preference. Polit Behav 29, 343–367 (2007). https://doi.org/10.1007/s11109-007-9030-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11109-007-9030-0