Measurement of implicit and explicit attitudes toward Barack Obama



The present study evaluated implicit and explicit attitudes toward Barack Obama in a student sample assessed during the 2008 election season. Implicit measurement was based on the Single Category Implicit Association Test (SC-IAT), in which participants categorized photographs of Obama as target objects associated with positive and negative evaluation categories. The relative darkness of Obama's image in the photographs was manipulated to accentuate racial cues. The results showed significant differences in implicit and explicit attitudes toward Obama between self-identified conservative and liberal students, as well as significant relationships between implicit and explicit measures and between implicit measures and modern racist attitudes. A stronger negative associational bias was found for conservative students, but not their liberal counterparts, to darker images of Obama in the implicit association test. The results underscore the importance of taking individual differences into account in measuring implicit responses to target stimuli. The results also support the convergent validity of implicit measures of attitudes toward a prominent political figure, as well as the construct validity of the SC-IAT. © 2010 Wiley Periodicals, Inc.

Perceptions of personality or character traits of political candidates play an important role in the electoral process (Funk, 1999; Renshon, 1996). The marketing of political candidates has become so much a part of election campaigns that candidates and their managers are often more concerned with projecting an appealing persona than with strongly promoting a particular ideology (Caprara & Zimbardo, 2004). Voters tend to develop simplified perceptions of political candidates during the course of election campaigns (Caprara, Barbaranelli, & Zimbardo, 2002). Campaign managers and public relations staffers use print and electronic media to frame a positive view of their candidates on traits such as trustworthiness, sound judgment, and honesty, while using negative advertisements against opponents. Voter perceptions in high-profile campaigns are molded by a constant barrage of political attack ads that attempt to reduce an opponent's character to a few undesirable traits, such as “wishy-washy,” “weak,” and “untrustworthy.” Evidence shows that perceptions of personality or character traits of politicians make independent contributions to voting behavior that cannot be explained simply by party affiliation (Pillai et al., 2003). They also are robust predictors of the public's approval of a sitting president (Greene, 2001).

The historic election of Barack Obama as the first African American president of the United States brought issues of race to the center of the American political process. During the campaign, media reports were replete with discussions about the effects that hidden racial bias might have on the election. One of the most frequently mentioned concerns was the so-called Bradley effect, whereby white voters openly express support for a black candidate but then fail to vote for the candidate in the privacy of the voting booth (Couzin, 2008). An AP–Yahoo News poll conducted shortly before the 2008 election raised concerns that racial bias might cost Obama the victory in a close contest, as negative stereotypes of Blacks as “lazy” and “violent” were endorsed by one-third of white Democrats (Racial Views, 2008). Although Obama won a decisive victory, the effects of underlying racial biases may have been mitigated by other pressing factors weighing on the minds of the voting public, especially the mounting economic crisis.

A major limitation of measuring attitudes toward political candidates is reliance on explicit measures, such as questionnaires, that may be subject to underlying response biases, such as social desirability responding. These concerns are magnified in the context of making judgments about minority candidates. People may fail to reveal underlying racial attitudes when questioned directly or may lack access to any underlying biases. More generally, the importance of measuring both explicit and implicit attitudes is that these measures may diverge because explicit measures are more subject to response biases. The advent of methods of measuring implicit attitudes provides a means of examining attitudes toward political candidates, including candidates of color, without relying solely on explicit measures.

Investigators recently reported that implicit measures of political attitudes were good predictors of future voting behaviors (Arcuri et al., 2008). Other investigators reported reduced levels of implicit anti-black attitudes in student samples obtained during the 2008 presidential election as well as indications that these reductions may have resulted from increased accessibility of positive black exemplars (Plant et al., 2009).


Evidence has accrued supporting the construct validity of implicit measures of racial attitudes as predictors of race-related attitudes and behaviors (Fazio & Olson, 2003; McConnell & Leibold, 2001; von Hippel, Brener, & von Hippel, 2008; Ziegert & Hanges, 2005). Implicit evaluative preferences favoring Whites over Blacks are found among both black and white Americans (Ashburn-Nardo, Knowles, & Monteith, 2003; Nosek, Banaji, & Greenwald, 2002).

Investigators have also begun to explore individual differences in implicit racial attitudes. For example, right-wing authoritarianism was linked to implicit racial biases as measured by the most widely used measure of implicit attitudes, the Implicit Association Test (IAT) (Rowatt & Franklin, 2004). Kemmelmeier (2004) also found higher levels of authoritarianism among supporters of Republican presidential candidates than supporters of Democratic candidates.

In other recent research, investigators compared four groups formed on the basis of obtaining either high or low scores on explicit and implicit measures of prejudiced attitudes (Son Hing et al., 2008). Modern racists (high on both explicit modern racism and implicit prejudice) were conservative and most strongly endorsed prejudice-related ideologies, whereas principled conservatives (high explicit modern racism, low implicit prejudice) held economically and politically conservative beliefs but did not strongly endorse prejudice-related ideologies.

Although explicit and implicit measures of racial attitudes tend to be correlated, the two sets of measures assess distinct sources of variance that cannot be accounted for by a single factor (Cunningham, Preacher, & Banaji, 2001). Measurement of implicit racial attitudes may also provide a means of detecting underlying racial attitudes that are not as subject to social desirability responding as more explicit measures (Fazio & Olson, 2003). Applied in the political arena, it would be useful to measure implicit measures of attitudes toward political candidates of color.


The physical characteristics of stigmatized groups, such as distinctive facial features and skin color, are important markers of race-based stereotyping. For example, Afrocentric features of both African American and European American target stimuli facilitate automatic stereotypic associations (Blair, Judd, & Fallman, 2004). The phenomenon of skin tone prejudice, or colorism, is another example of physiological features activating stereotyped associations. Minority group members with darker skin tones are subject to more negative stereotyping by both majority and minority groups than their lighter-skinned counterparts (Maddox & Gray, 2002; Uhlmann et al., 2002).

Prior research shows that physical features of political candidates influence perceptions of personality traits. In one study, for example, subtle alterations of the facial features of several U.S. presidents—making them appear more mature or neotenous—affected social perceptions of their power and warmth (Keating, Randall, & Kendrick, 1999). More recently, baby-facedness was found to be negatively related to perceptions of competence of politicians based on photograph-based judgments of 1785 Finnish political candidates (Poutvaara, Jordahl, & Berggren, 2009).


Although prior research on implicit attitudes demonstrates implicit racial biases toward African Americans in general, no reported studies have examined implicit attitudes toward a highly esteemed African American target person. Prior research on perceptions of political candidates has also been limited to explicit measures. To our knowledge, this was the first attempt to measure implicit attitudes toward a leading political figure.

The present study extends prior research by examining both explicit and implicit attitudes toward a high-profile minority politician, Barack Obama. We also incorporated a skin-tone manipulation by lightening or darkening the same photographs of Mr. Obama used in a test of implicit associations. The study design incorporated two between-subjects variables, Skin-Tone condition (lighter vs. darker target photos of Mr. Obama) and self-identified Political Ideology (conservative vs. liberal). Based on prior research supporting the role of individual differences variables in implicit attitudes, we hypothesized that self-identified liberal students would show more positive implicit and explicit attitudes toward Mr. Obama than their conservative counterparts. Second, we expected to find more favorable implicit and explicit attitudes toward Mr. Obama to be associated with lower levels of endorsement of modern racist attitudes. Finally, we expected that darker skin-tone coloring of target photographs of Mr. Obama would yield more negative implicit attitudes than lighter skin-tone coloring.



Seventy-eight introductory psychology students (22 males, 56 females; 81% freshmen and sophomores) drawn from introductory psychology courses in a large, urban, Catholic university in the northeastern United States during the fall 2008 semester participated in the experiment. The racial/ethnic breakdown of the participants was as follows: African American (21), Caucasian (19), Asian (12), Hispanic/Latino (13), multiethnic (11), and other/missing (2). Participants received credit in partial fulfillment of a course requirement in return for their participation.

Explicit Measures

Participants provided ratings of Mr. Obama on a widely used measure of perceived personality traits of political candidates, the Big Five Questionnaire (BFQ; Caprara et al., 1993). The BFQ consists of the following 25 adjectives that serve as markers of the five-factor model: self-assured/confident, persistent/persevering, poised/composed, cordial/warm, energetic/lively, cheerful/happy, generous/kindhearted, sincere/genuine, truthful/believable, loyal/dutiful, responsible/conscientious, reliable/dependable, peaceful/calm, stable/anchored, optimistic/hopeful, confident/upbeat, smart/intelligent, innovative/original, creative/imaginative, enterprising/resourceful, precise/exacting, active/dynamic, inventive/ingenious, modern/contemporary, and efficient/skillful. In order to reduce the number of variables in the analysis, we used two factorially derived dimensions derived from the 25 adjective ratings, which were labeled energy/innovation and honesty/trustworthiness (Caprara, Barbarenelli, & Zimbardo, 1997).

Using individual 5-point rating scales, participants also rated Mr. Obama on the following traits commonly associated with perceptions of political candidates: trustworthiness, leadership potential, likeability, and warmth. For purposes of analysis, a composite score was formed based on the average ratings across these four traits, which yielded a Cronbach's alpha of 0.90 on the composite scale. Participants also provided an overall evaluation of Mr. Obama on a 9-point bipolar scale ranging from “extremely negative” to “extremely positive.”

A six-item scale measuring racist attitudes arising from conflicts about the civil rights movement, the Modern Racism Scale (McConahay, 1986), was used as an explicit measure of racial attitudes. Respondents used a 5-point Likert scale ranging from (1) “strongly disagree” to (5) “strongly agree” to reflect their degree of endorsement of modern racist beliefs, such as beliefs that Blacks have received more economic benefits and more favorable treatment from government and news media than they deserve. Prior evidence shows significant associations between implicit racial attitudes and scores on the Modern Racism Scale (Cunningham, Preacher, & Banaji, 2001).

The individual difference variable, Political Ideology, was measured by means of a single, 7-point bipolar scale with scale points ranging from “conservative” to “liberal.”

Implicit Measures

The Single Category Implicit Association Test (SC-IAT) (Karpinski & Steinman, 2006) was used to measure implicit attitudes toward Mr. Obama. The SC-IAT is a modification of the widely used Implicit Association Test (IAT) (Greenwald, McGhee, & Schwartz, 1998), which measures strength of association based on a single target object rather than the two targets required by the standard IAT. The SC-IAT was used in the present study because the target object comprised photographs of a single individual, Barack Obama, rather than a dual category comprising photographs of Blacks and Whites.

The SC-IAT is based on the assumption underlying the IAT that categorization of target objects will be easier and thus faster when they are paired with an evaluative category consistent with the participants' underlying attitudes than when they are paired with an incongruent category (Son Hing et al., 2008). Evidence supports both the reliability and validity of the SC-IAT (Karpinski & Steinman, 2006). A recent meta-analysis of the IAT supported its predictive validity and, importantly, showed greater predictive validity of the test in comparison to self-report measures in predicting black–white interracial behaviors (Greenwald et al., 2009).

The target objects consisted of a set of five color photographs of Barack Obama obtained from an Internet search site. The photographs were all head shots, 1.5″ (width) × 2″ (length) in size, which were projected on a computer screen and cropped to eliminate surrounding details. A photo editing program,, was used to create lighter and darker versions of each photograph. The degree of adjustment needed to create the lighter and darker versions ranged from a 16.5% increase to a 25% decrease depending on the lightness of the original photographs. The target objects were paired with positive and negative evaluative categories to obtain measures of latency of response (in milliseconds) in performing the categorization task.


Participants first completed the implicit association test, followed by completion of the explicit measures. Using the SC-IAT protocol, participants were individually tested while seated at a desk in front of a desktop computer in a private office. The SC-IAT was presented on a computer screen utilizing the Inquisit program. Participants were randomized to two skin-tone SC-IAT conditions in which the same photographic images of Obama were used as target objects but differed in relative lightness of skin tone.

After a general introduction to the SC-IAT task, participants were asked to categorize positive words on the left side of the screen (using the “e” key) and negative words on the right side of the screen (using the “i” key). The positive word set used in the Inquist program consisted of the words happiness, peace, joy, flower, and love; the negative word set consisted of the words death, pain, murder, hatred, and war.

Participants first completed 20 practice trials in which they were asked to categorize positive and negative target words that were presented in the center of the screen. The experimental phase comprised two test blocks of 72 trials each that were counterbalanced across participants to control for order effects (i.e., Obama + positive and Obama + negative). After completion of the SC-IAT, participants were administered the explicit attitude measures and provided general demographic information. The SC-IAT showed a reliability coefficient in the present study of 0.76 based on the Spearman–Brown correction for split-half reliability.

SC-IAT scores were computed based on the procedures of Karpinski and Steinman (2006). Each trial yielded response latencies in milliseconds; only trials from the two test blocks were included in the analysis. All trials resulting in an error (e.g., depressing e for war) prompted a replacement of the original response latency with an adjusted score of 400 milliseconds above the participant's block mean. The difference or D-score was derived by subtracting positive latencies from negative latencies, then dividing by the standard deviation of all correct responses from the two test blocks.


Implicit Attitudes Toward Barack Obama

A preliminary analysis was conducted on comparisons of D-scores of dummy-coded student groups based on racial/ethnic self-identification: Blacks versus Caucasians, Blacks versus Hispanics/Latinos, and Blacks versus Asians. Blacks were selected as the default category in these comparisons since the study concerned measurement of implicit attitudes toward a high-status African American target person. None of these preliminary analyses showed significant differences between Blacks and other groups (ps > 0.05). We obtained an overall D-score of −0.025 (SD=0.39), showing a negligible negative associational bias toward Mr. Obama among our student participants.

Consistent with predictions, a test of the means for D-scores for self-identified conservative and liberal student groups showed a difference of moderate magnitude [t(56)=2.41, p < 0.05, d=0.69]. The mean D-scores were 0.085 (SD=0.26) for liberals and −0.183 (SD=0.59) for conservatives, indicating a slightly positive implicit bias among liberals and a slightly negative implicit bias among conservatives.

Regression Analysis of Implicit Associations Based on Political Ideology and Skin-Tone Manipulation

Participants were divided into conservative and liberal groups based on their responses to the conservatism/liberalism scale. Participants scoring below the midpoint on the scale (n=38) were classified as conservative and those scoring above the midpoint (n=20) were classified as liberals. In the effort to limit the analysis to self-identified conservatives and liberals, we eliminated participants (n=20) who endorsed a midpoint value (i.e., “fence-sitters”) on political ideology.

Hierarchical multiple regression was performed on Winsorized D-scores by entering two predictor variables, Skin Tone (light vs. dark) and Political Ideology (conservative vs. liberal). Main effects were entered in the first step of the analysis, which was followed on the second step by entering main effects and the interaction effect. The regression equation in the first step was marginally significant (p=0.06) and was accounted for by Political Ideology [β=−0.29, t(56)=2.1, p < 0.05], showing, as expected, a more negative associational bias toward Mr. Obama among conservative students overall.

The full regression model at step 2 was significant [R2=0.17, F(3,54)=3.6, p=0.02], with the interaction term emerging as the only significant predictor [β=−0.51, t(55)=−2.1, p < 0.04]. Although the main effect of skin-tone alteration was not significant, the relationship is best explained by the higher-order interaction effect of Skin Tone × Political Ideology. As shown in Figure 1, the decomposition of the interaction shows that the skin-tone manipulation more strongly influenced associational bias among conservative students. Conservative students exposed to the lightened image of Mr. Obama showed a slight positive associational bias with respect to D-scores (M=0.08); those viewing the darkened photos showed a slight negative bias (M=−0.18).

Figure 1.

Implicit attitudes toward Barack Obama.

Tests of specific main effects showed a marginally significant effect representing a more negative associational bias toward Mr. Obama, on the average, among conservative students in the darker skin-tone condition than in the lighter condition [t(18)=1.9, p=0.076, d=0.87]. By contrast, skin-tone manipulation failed to affect implicit associations among liberal students [t(36)=0.83, p=0.41]. Moreover, although liberal and conservative students did not differ in implicit responses in the lighter skin-tone condition (p > 0.05) they did differ significantly (and substantially) in the darker skin-tone condition with conservative students showing a more negative associational bias [t(27)=2.99, p < 0.01, d=1.01].

Analysis of Explicit Attitudes

A MANOVA was performed to examine differences between conservatives and liberals on explicit attitudes, as measured by BFQ ratings, composite score of positive traits, and overall evaluation, as well as modern racist attitudes. A significant multivariate effect was obtained [F(5,52)=9.661, p < 0.001, partial η´2=0.48]. As hypothesized, follow-up between-subjects tests showed that liberals and conservatives differed significantly in their views of Mr. Obama (i.e., more positive ratings of Obama among liberals) on the BFQ honesty/trustworthiness factor (F=16.41, p < 0.001, partial η´2=0.23), composite traits (F=11.37, p < 0.001, partial η´2=0.17), and overall evaluation (F=31.21, p < 0.001, partial η´2=0.34), whereas the BFQ energy/innovation factor exhibited a near significant effect in the same direction (F=3.94, p=0.052, partial η´2=0.07). Liberals and conservatives also differed significantly with respect to modern racist attitudes, with liberals showing lower scores on this scale than conservatives (F=6.38, p=0.01, partial η´2=0.10).

Correlations among Explicit and Implicit Measures

Table 1 shows intercorrelations among implicit and explicit measures of attitudes. D-scores were significantly correlated with each of the explicit measures with the exception of energy/innovation. Significant correlations between implicit and explicit measures ranged from 0.26 to 0.37, suggesting that these measures, though interrelated, were relatively distinct from each other. As predicted, implicit attitudes (D-scores) toward the target object of Mr. Obama were also related to modern racist attitudes in the expected direction [r(76)=−0.37, p < 0.01], providing support for the construct validity of the implicit measures with respect to predicting race-related attitudes.

Table 1. Correlations Among Explicit and Implicit Attitudes (N=78).
  • *

    * p < 0.05;

  • **

    ** p < 0.01.

1. D-Score       
2. Energy/innovation0.930.14     
3. Honesty/trustworthiness0.950.26*0.86**    
4. Modern racism0.79−0.37**−0.49**−0.57**   
5. Composite traits0.900.30**0.76**0.86**−0.70  
6. Overall evaluation 0.36**0.70**0.82**−0.60**0.85** 


The 2008 election of Barack Obama to the nation's highest office was a watershed event in American society. However, despite the election of the nation's first African American president, few would argue that we have transcended issues of race in our society. Although Americans today may no longer endorse the flagrant racist beliefs that permeated American society in the Jim Crow era of racial segregation, more subtle forms of racism arising from perceptions that Blacks in the United States have received favorable treatment as a result of the civil rights movement continue to affect race-related attitudes and behaviors.

The present study examined implicit and explicit attitudes toward Barack Obama and the relationships between these attitudes and modern racist beliefs. The study design also included a manipulation of skin-tone of the target images of Mr. Obama to examine whether accentuated racial cues would affect implicit associations.

Not surprisingly, we found that self-identified liberal students showed more positive explicit attitudes toward Mr. Obama than their more conservative counterparts. We also found similar differences in implicit attitudes toward Mr. Obama, with liberal students showing a small positive associational bias and conservative students showing a small negative associational bias. However, these differences were subsumed by a higher-order interaction of Skin Tone × Political Ideology. Specifically, conservative and liberal student groups did not differ significantly in their implicit responses to Mr. Obama when his skin tone was lightened, but did differ significantly (conservatives showed more negative responses) when his skin tone was darkened. Moreover, liberal students showed a slight positive associational bias in both the darker and lighter skin-tone conditions. Although conservative students had a slight positive associational bias to Mr. Obama in the lighter skin-tone condition, they showed a more pronounced negative bias when responding to the darkened images of Mr. Obama. A large size of effect was found for the difference between these conditions for conservative students. These results are potentially significant to political consultants and marketers, as they illustrate that variations in skin tone can have a marked effect on implicit responses in certain target groups but not in others. The results also point to the importance of taking individual differences variables into account when measuring implicit responses of target groups.

Despite the small magnitudes of implicit associational bias, evidence of convergent validity of implicit measures was demonstrated by significant interrelationships between implicit and explicit attitudes. Moreover, the construct validity of the implicit measures was strengthened by findings of predictable relationships with a measure of modern racist attitudes.

The potential applications of the implicit measures in marketing research extend well beyond marketing of political candidates. The advent of methods of measuring implicit attitudes provides investigators in marketing research with a new window through which to examine automatic associations to a range of product-related stimuli and, as represented in the present case, high-profile political candidates. The present findings underscore the importance of considering individual differences variables in measuring implicit responses to product cues and advertising messages. Marketers may benefit from examining the role of potential moderators of implicit responses, such as demographic factors (gender, race, age, education), history of product use, product preferences, attitudinal dimensions, and personality traits. Implicit measurement may also reveal underlying racial biases and how they may come into play in determining perceptions of high-profile persons of color, including entertainers, sports figures, and other political leaders.

Several limitations of the present study should be noted. The sample was restricted to a college sample comprised of mostly freshman and sophomore introductory psychology students, and results may not generalize to other groups. Second, although a measure of race-related attitudes was included in the study and showed predictable relationships with implicit measures of attitudes toward Mr. Obama, the predictive value of implicit attitudes would be further strengthened by demonstrating relationships with more direct race-related behaviors. Moreover, the effects of skin-tone darkening as a cue for eliciting stereotypical biases may have been mitigated by using such a high-prestige and well-known stimulus object as Barack Obama. Future research on implicit skin-tone prejudice may benefit from using black and white faces of anonymous individuals as stimulus objects. Another avenue of future research would be to more closely examine attitudes of prospective voters representing the middle of the political spectrum. It is conceivable that other moderator variables apart from political ideology might influence implicit associations to political candidates.

It is conceivable that initial measurement of implicit attitudes may have influenced responses to the explicit measures. Although all subjects completed the measures in the same order to control for any confounding effects across levels of skin-tone manipulation and political ideology, it is conceivable that priming implicit attitudes may have affected subsequent measurement of explicit attitudes toward Mr. Obama.

The advent of methods of measuring implicit attitudes provides investigators in marketing research with a new window through which to examine automatic associations to a range of product-related stimuli and, in the present case, high-profile political candidates. Implicit measurement also may reveal underlying racial biases and how they may come into play in determining perceptions of different ethnic or racial groups and political candidates who are members of these groups.