racial stereotypes, racial context, and the 2008...

45
Racial Stereotypes, Racial Context, and the 2008 Presidential Election * Jason H. Windett Assistant Professor Saint Louis University [email protected] Kevin K. Banda Ph.D Candidate The University of North Carolina at Chapel Hill [email protected] Thomas M. Carsey Thomas J. Pearsall Distinguished Professor of Political Science The University of North Carolina at Chapel Hill [email protected] September 4, 2013 Abstract As the first African-American nominee for President of a major political party, Barack Obama’s campaign and ultimate victory reminded voters, scholars, pundits, and the press of the centrality of race in American political life. Speculation by observers of all types centered around the potential impact of race as an individual psychological prejudice and/or as a geographic/contextual factor. These two themes parallel different leading scholarly treatments of race and racism in the U.S. Rather than choose one theme or the other, in this paper we bring both traditions together in a unified analysis of white voter response to Obama. We find strong evidence that the level of prejudice toward African-Americans held by whites affected their evaluations of Obama as well as their probability of voting for him. In contrast, we find little evidence that whites responded to the racial context of their immediate geographic environment. * Forthcoming in Politics, Groups, and Identities. Please direct all correspondence regarding this submission to [email protected]. 1

Upload: others

Post on 12-Aug-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Racial Stereotypes, Racial Context, and the 2008Presidential Election∗

Jason H. Windett†

Assistant ProfessorSaint Louis [email protected]

Kevin K. BandaPh.D Candidate

The University of North Carolina at Chapel [email protected]

Thomas M. CarseyThomas J. Pearsall Distinguished Professor of Political Science

The University of North Carolina at Chapel [email protected]

September 4, 2013

Abstract

As the first African-American nominee for President of a major political party,Barack Obama’s campaign and ultimate victory reminded voters, scholars, pundits, andthe press of the centrality of race in American political life. Speculation by observersof all types centered around the potential impact of race as an individual psychologicalprejudice and/or as a geographic/contextual factor. These two themes parallel differentleading scholarly treatments of race and racism in the U.S. Rather than choose onetheme or the other, in this paper we bring both traditions together in a unified analysisof white voter response to Obama. We find strong evidence that the level of prejudicetoward African-Americans held by whites affected their evaluations of Obama as wellas their probability of voting for him. In contrast, we find little evidence that whitesresponded to the racial context of their immediate geographic environment.

∗Forthcoming in Politics, Groups, and Identities.†Please direct all correspondence regarding this submission to [email protected].

1

Page 2: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Race has defined political conflict for much of the United States’ history. It presented

a Constitutional crisis at the country’s birth, nearly tore the country apart through civil

war, and re-emerged beginning in the 1960’s as a fundamental cleavage in American party

politics. As an African American, Barack Obama’s successful campaign for the Presidency

in 2008 against a white opponent, John McCain, renewed the discussion of race in American

politics. It also presents scholars with a historic opportunity to explore the impact of race

in electoral politics as voters faced, for the first time, a choice between an electorally viable

African American candidate and a white candidate in the general election for President.

While several commentators and political pundits argued that the Obama victory marked

the beginning of a “post-racial” society in the United States, early analyses of the impact

of Obama’s race on the 2008 election have provided mixed results. For example, Mas and

Moretti (2009) found little evidence that Obama performed differently due to his race com-

pared to John Kerry in 2004 and Congressional Democrats in 2008. In contrast, Ansolabehere

and Stewart (2009) argued that Obama’s race led to an increase in turnout among the black

population, which contributed to his victory.

More nuanced examinations of voting behavior revealed a backlash against Obama among

white voters driven by their racial attitudes. Using an experimental design, Schaffner (2011)

examined the impact of racial salience on vote choice and found that Obama’s race led to

a three percentage point decline in his share of the white vote. Parker, Sawyer and Towler

(2009) found that symbolic racism contributed to evaluations of Obama, though this finding

is produced from data collected in a single state. Lewis-Beck, Tien and Nadeau (2010)

argued that racial resentment felt towards Obama displaced the effect of economic voting

among some voters, costing Obama what could have been a landslide victory.

In the most expansive analyses of the 2008 election, both Piston (2010) and Tesler and

Sears (2010) examined the direct impact of racial prejudice of white respondents on their

2

Page 3: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

vote choice. These works use the ANES questions on racial attitudes towards blacks in

order to evaluate the likelihood of voting for Obama. Using data from 1992 through 2008,

both works conclude that the 2008 election is the only presidential election in which racial

sentiments predicted the behavior of white voters. Tesler and Sears (2010, 6) note that “the

two sides of racialization — that is, the racially resentful opposition to and racially liberal

support for Obama — resulted in a considerably larger influence of racial attitudes on the

presidential vote in 2008 than any other campaign in modern history.” Block Jr. (2011)

supports this argument and shows that race actually affected white voters at greater levels

than black voters. He finds that black voters were considerably less likely to emphasize race

when casting their ballots for Obama compared to white voters who voted against Obama.

These findings follow in a long line of research exploring race and voting behavior in the

United States. While few deny the centrality of race in the American political experience,

there remains a surprising degree of uncertainty over how best to understand it. There are

at least two major approaches in the existing literature. One tackles the problem of race

as an individual-level attitude or opinion held by members of one group toward members

of another. The out-group is defined in collective terms, but the engine for a race-based

response lies within the individual’s internal formation of feelings of racial prejudice. A

number of differences exist among scholars working in this general area, but they share in

common an internal psychological response to race.

A second approach views race as an external contextual factor that triggers a response

among members of a particular group. The notion is that members of one group (e.g. whites)

will behave differently depending on the relative size of the population of another group (e.g

blacks) in their surrounding area. With its origins in Key’s 1949 “racial threat” hypothesis,

the critical causal mechanism in this perspective exists outside the individual in the form of

the racial make-up of someone’s surrounding area.

3

Page 4: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

These two perspectives are not absolutes nor unrelated. Certainly the racial threat hy-

pothesis rests on an individual reaction to the surrounding context, while psychological

approaches clearly define racism in terms of how individuals respond to other groups. How-

ever, each perspective points to a different locus for racial politics — internalized racism

or external contextually-defined “threats.” As we will show, there are a number of ways to

conceive of how these two perspectives might be integrated into a single model of candidate

evaluation. We employ direct measures of both racial context and racial attitudes in this

research to predict how survey respondents evaluated the two major party candidates for

President as well who they reported voting for in the 2008 election.

Approaches to Race in American Politics

Individual-level Racial Prejudice

The literature on racism within the United States is vast and well-beyond comprehensive

review here. In this section, we limit our focus to the individual-level phenomenon of racial

prejudice as manifest in the ascription of racial stereotypes to African-Americans by whites.

The use of stereotypes to measure prejudice has a long history in the social sciences.1

We first consider the degree to which racial hostility towards African-Americans as a

group persists among whites in the U.S. Sniderman and Carmines (1997) state, “Every

systematic study of long-term trends in American racial attitudes, without exception, has

concluded that race prejudice has dramatically declined since 1940” (p. 65). Of course, the

decline in racial prejudice does not mean that it has disappeared. Using responses of whites

surveyed as part of the 1991 Race and Politics Survey, Sniderman and Carmines (1997) show

that surprisingly large numbers of whites are unwilling to ascribe positive traits to blacks as

1Interested readers should review the citations provided in Levine, Carmines and Sniderman (1999)regarding the study of stereotypes.

4

Page 5: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

a group. They further report that anywhere from 25 to 50 percent of white respondents are

willing to ascribe negative traits such as “lazy,” “boastful,” “irresponsible,” “complaining,”

or “aggressive or violent” to blacks as a group (pp.62-3).

Advocates of the symbolic racism perspective argue that the contemporary political cli-

mate leads respondents to provide socially desirable answers to survey questions rather than

directly express racial prejudice (e.g. Kinder and Kiewiet 1981; Tarman and Sears 2005).

This possibility raises some question as to the accuracy of direct survey-based questions.

However, Sniderman and Carmines (1997) convincingly demonstrate through survey-based

experiments that respondents who express positive or negative views of blacks through stereo-

types are accurately reflecting their underlying beliefs. Furthermore, Levine, Carmines and

Sniderman (1999) report that the willingness of whites to apply positive or negative stereo-

types to African-Americans is largely structured by a single dimension, lending credence to

the idea that such stereotype questions tap into a single underlying racial prejudice dimen-

sion.

To what degree do feelings of racial prejudice shape the political attitudes and evaluations

of whites? Citrin, Green and Sears (1990) argue that racial attitudes affect voters’ evalu-

ations of candidates, explicitly racial policies, and even policies that lack an explicit racial

component. More recently, Sniderman and Carmines (1997) show that racism significantly

predicts the views whites hold on a number of policies designed to benefit African-Americans,

but that the impact of racial prejudice was frequently substantively smaller than they ex-

pected.

Race as a Contextual Factor

Key’s 1949 classic study of Southern politics gave birth to the contextual study of race

in American politics both in the South (e.g. Black and Black 1987) and across the country

5

Page 6: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

as a whole (e.g. Huckfeldt and Kohfeld 1989). Key’s central contribution in this field has

been termed the “racial threat” hypothesis. Key found that the willingness of whites to

engage in political actions viewed as hostile to the interests of blacks depended on the size

or density of the black population in the surrounding area. The racial threat hypothesis

assumes that whites have political interests that differ from blacks and that larger black

populations represented a political threat to what would otherwise be white dominance of

politics. Thus, the real threat depended on how government would be expected to respond

if African-Americans achieved meaningful political influence. In this regard, Key essentially

argued that whites viewed the potential political threat of blacks as a numbers game.

Key’s original thesis has received substantial support both in and out of the South (e.g.

Wright 1977; Huckfeldt and Kohfeld 1989). Glaser (2003), for example, reports evidence from

experiments showing whites as less supportive of policy benefits being distributed evenly

among residents of an area as the proportion of the population in the area that is black

increases.

Others, however, have found a positive rather than negative response among white voters

to increases in black population densities. For example, Carsey (1995) finds that white

support for black mayoral candidates in New York and Chicago increased as the size of the

black population increased in a precinct. Liu (2003) finds a deracialization effect among

white voters living in majority black areas. Others debate the true effect of racial context on

how whites respond (see for example Giles and Buckner 1993, 1996; Voss 1996), or present

mixed results (e.g. Voss and Lublin 2001). Campbell, Wong and Citrin (2006), for example,

find only limited support for racial contextual effects on white voter behavior regarding three

racially charged ballot initiatives in California.

Oliver and Mendelberg (2000) offer a detailed explanation of the contextual determinants

of racial attitudes and conclude that these attitudes are conditional on a number of variables.

6

Page 7: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

First, they observe the overly simplified nature of the earlier analyses on the racial threat

hypothesis. They argue that the contextual effects of race are deeply intertwined with

socioeconomic status. Second, they conclude that the level of analysis is also an important

consideration when analyzing contextual effects. They show no relationship between racial

attitudes and racial composition at the zip code level, but when they expand the scope of

geography to a larger geographic space, racial antagonism is moderately related to the size

of the black population. Hopkins (2010) offers a similar conditional argument for the affect

of the size of nearby immigrant populations on attitudes regarding immigration policy.2

Carsey and Windett (2013) suggest that these conditional—or what might be better

called mixed—findings may result from differences in the expectations whites hold regarding

a perceived “threat” from black political power compared to whites’ actual experiences with

black political leaders. The absence of real experience may lead voters to rely on race as

a low information cue that might shape how they vote (Williams 1990; Terkildsen 1993;

McDermott 1998). Hajnal (2001) argues that actual experience with black city officials led

whites to be more likely to vote for their re-election. However, Carsey and Windett (2013)

find little evidence that prior experience with a black elected official influenced the voting

behavior of whites in five statewide elections held in 2006 that involved an African-American

candidate.

A Unified Approach to Race and Racism

The brief reviews provided in the previous two sections highlight three features of our

current understanding of racial politics in the U.S. First, there exists evidence of a nega-

tive response among whites toward African-Americans at both the individual-level and the

2Blalock (1967) offers a slightly different, though essentially conditional, argument, suggesting that theimpact of the size of a minority population on the views held by others might be non-linear. We exploredthat possibility in this analysis and found no evidence to support it.

7

Page 8: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

contextual-level. Whether or not prejudice is on the decline, it remains a potentially pow-

erful force in American politics. Second, empirical results are mixed, particularly regarding

the contextual effect of black population density on the behavior of whites. A multitude

of explanations have been offered for these mixed findings (see Carsey and Windett 2013

for a review), but the more general conclusion is that we do not yet understand how racial

contexts become politically meaningful for individuals immersed in them. Third, vast liter-

atures exist in these two areas of research, but comparatively less has been done to examine

both individual and contextual factors together.

Missing from our discussion thus far is how race becomes politicized in a particular cir-

cumstance. Mendelberg (2001) rightly notes that how, and to what degree, race is activated

as a political issue for a given electoral contest fundamentally shapes how voters will respond.

She finds that subtle appeals to racial animosities might stimulate a negative reaction among

white voters to black interests, but that explicit racist appeals may actually generate sym-

pathy rather than fear among whites (but see Hutchings, Walton Jr. and Benjamin 2010).

Sears et al. (1997) also suggest that the manner and extent to which the underlying racial

prejudices held by many whites emerge depends on whether and how they are primed. Simi-

larly, Hopkins (2010) finds that the impact of proximity to immigrant populations on citizen

attitudes toward immigration depend on the combination of a sudden change in the size

of the proximate immigrant population and whether national political rhetoric regarding

immigration is particularly sharp and prominent.

Black candidates themselves rarely seek to inject race as an issue in their electoral cam-

paigns (Lupia and McCubbins 1998). As a result, the importance of race to white voters

may not differ substantially when a black candidate is running compared to when one is not

(e.g. Citrin, Green and Sears 1990). However, the emergence of race was largely unavoidable

in the first general election contest for president involving an African-American candidate

with a real chance to win. Race was firmly injected into the Presidential contest leading up

8

Page 9: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

to the South Carolina Democratic primary, but press coverage of the primary and general

election campaigns likely made race unavoidably salient. Bell (1997) believes that white

negative backlash is unavoidable for black candidates, but Hajnal (2001) argues that white

voters learn from experience with black candidates and elected officials in a way that reduces

their negative fears of black political leaders.

We also consider the possibility that individual racial prejudice may be shaped by the

racial context within which individuals live. Key’s racial threat hypothesis may be broadened

to predict that negative views of blacks more generally will emerge among whites who live

in areas with large black populations. In contrast, Allport (1954) suggests that greater

interaction with individuals from another group would serve to erode prejudice and foster

the development of a shared identity with common goals. Carsey (1995) suggests such a

mechanism may be at work in explaining his finding of an increased likelihood of white

voters voting for a black candidate for mayor as the percentage of the population in a white

respondent’s neighborhood increased. Finally, Kinder and Mendelberg (1995) find that racial

context conditions the impact of racial prejudice on attitudes regarding race-based policies,

with greater proximity to racial minority populations lowered the impact of prejudice on

policy attitudes.

In sum, the existing literature points to a number of different models that might explain

how both racial context and racial prejudice might contribute to explaining how people

evaluate candidates for a given election. The simplest model is a basic additive model, such

as the one represented in Equation 1, where candidate evaluation (CE) is expressed as a

simple linear additive function of racial context (C) and racial Prejudice (P):3

CE = β0 + β1(C) + β2(P ) + ε (1)

3Again, recall that Blalock (1967) would suggest including both C and C2.

9

Page 10: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

The theoretical motivation for a model like Equation 1 assumes that the separate arguments

made regarding the impact of racial prejudice and the impact of racial context are essentially

independent of each other. Correlation between racial attitudes and racial context might

exist, of course, but correlation among independent variables is a normal common feature of

regression models.

If context has a direct effect on candidate evaluation, but also impacts racial prejudice,

then the resulting 2-equation model would captured by Equations 2 and 3:

P = α0 + α1(C) + ε1 (2)

CE = β0 + β1(C) + β2(P ) + ε2 (3)

This model rests on the assumption that racial context is an exogenous variable, but that

racial attitudes are endogenous to context as well as being a potentially intervening variable

between context and candidate evaluation. If the sole reason that racial context impacts

candidate evaluation is because racial context drives racial prejudice, and that racial prej-

udice affects candidate evaluation, then we would expect α1 to be statistically significantly

different from zero in Equation 2, β2 to be statistically significantly different from zero in

Equation 3, but β1 in Equation 3 to not be statistically significantly different from zero.4

A third model suggested by the literature is that racial context and racial prejudice inter-

act with each other to impact how citizens evaluate candidates. That implies the following

model, represented in Equation 4:

CE = β0 + β1(C) + β2(P ) + β3(C × P ) + ε (4)

Such a model is consistent with Hopkins (2010) and Kinder and Mendelberg (1995), who

4Whether α1 and β2 would be positive or negative would, of course, depend on how the various variablesin the model were coded.

10

Page 11: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

argue that the impact of attitudes about racial or ethnic groups on some evaluation or opinion

is conditioned by context. Making statements about exactly what the coefficient estimates

would look like in a model that includes interactions is unwise because that depends on how

each variable is measured, but we can say that we should observe an improvement in model

fit moving from Equation 1 to Equation 4, and that we should see a meaningful change in

the marginal effect of racial prejudice on candidate evaluations as racial context changes. Of

course, we should see a similar meaningful change in the marginal effect of racial context on

candidate evaluations across the range of racial prejudice if Equation 4 was the appropriate

model.

In short, there is some theoretical foundation and at least partial empirical support for

each of these models in the existing literature. The bulk of the literature focuses on just the

impact of racial context or the impact of racial prejudice on candidate evaluations, policy

preferences, or some other attitude or behavior. In that sense, the majority of existing

studies do not even fully specify the model represented by Equation 1. Some studies have

examined limited versions of the other two models, but, to the best of our knowledge, the

literature has not previously made each of these competing models explicit and then tested

them against observed data.

In this paper, we argue that racial context and individually held racial stereotypes must

be brought together to understand how race might affect the views people hold regarding

candidates in a given election. We suspect that both might have direct effects on how citizens

respond when race is made salient in a particular election, which leads us to focus on analyses

based on Equation 1. However, we also explore whether contextual factors shape the racial

views people hold, as implied by Equation 2. We find no evidence for the model described by

Equation 4, so we do not discuss that model further.5 We evaluate these models using two

different data sets based on voter surveys conducted during the 2008 Presidential election.

5We also find no empirical support for including both our context measure and context squared.

11

Page 12: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Data and Methods

We explore the impact of individual-level racial prejudice and racial context on white

citizens in the context of the 2008 Presidential election using two nationally representative

surveys. Our data come from the 2008 Cooperative Congressional Elections Study (CCES)

and the 2008 American National Election Survey (ANES). The CCES surveyed about 33,000

respondents nationally as part of a multi-team effort. The survey itself was administered

online to a sample of opt-in respondents who were chosen to be representative of the national

population using a matching algorithm.6 The specific sample from which we run our analyses

consists of a subsample of 1,000 respondents who answered a series of unique questions. The

ANES survey contains over 2,100 nationally representative respondents.7

Individual Level Measures of Racial Prejudice

One of our primary explanatory variables of interest is our measure of the prejudicial

attitudes that individuals hold towards African-Americans as a group. We use three separate

but related measures of racial prejudice. Two of the measures are the degree to which

respondents reported believing that blacks are “hardworkers.”8 This is measured on a seven

point scale in the ANES data and a four point scale in the CCES data. Higher values

indicate more positive views towards blacks in both data sets. The mean and standard

deviation in the ANES data are 3.9 and 1.3. In the CCES data, these values are 2.9 and

0.8. Accounting for the different scales, this suggests that whites in the CCES sample on

average held more positive views towards blacks than did white ANES respondents. While

6The CCES collected data between October 8 and November 3 for the preelection battery and fromNovember 5 to December 1 for the postelection battery. The within-panel response rate for the 2008 CCESwas 47.1%.

7ANES data was collected from September 7 through November 3 for the preelection battery and betweenNovember 5 and December 21 for the postelection battery. The within-panel response rate was 66.3% andthe minimum response rate was 59.5%.

8Interested readers can find the question wording for these instruments in the Appendix.

12

Page 13: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

the range of these scales differ from one another, the measures themselves tap into the same

underlying concept. This allows us to compare results across data sets.9

We construct a third and more conceptually rich measure of racial prejudice through the

use of factor analysis on four separate indicators of attitudes towards blacks included in our

CCES data. Using a subset of stereotypes like those employed by Sniderman and Carmines

(1997), we asked respondents the degree to which they agreed that African-Americans are

“hardworking,” “trustworthy,” “violent,” and “complainers”10. Our resulting factor score

ranges from -2.26 to 1.18; lower scores indicate that respondents hold more negative stereo-

typical views of blacks while higher scores suggest that respondents hold more positive views

of blacks. This measure has a mean score of -0.47 and a standard deviation of 0.773. Be-

cause lower values on each score are associated with more negative views towards blacks

while higher values are associated with more positive views of blacks, the literature cited

above would predict that our measures would be positively related to a higher likelihood of

voting for Obama and an increased likelihood of holding positive views toward him.

Racial Context

We measure racial context as the black population density of each white respondent’s

environment by matching each respondent’s five digit FIPS county code to the percentage

of residents in the county who were black as reported in the 2000 Census. We measure

racial context at the county level for two reasons. First, the county level provides a unit

of geography that is small enough to capture the potential for interaction between whites

and blacks while also being large enough to extend well beyond direct interpersonal contact.

The racial make-up of a county is something residents are likely to be casually aware of, but

counties are large enough to allow for social segregation within the county and a possible

9This particular measure of racial attitudes is the only one that is included in both data sets.10See the Appendix for wording and coding of these questions.

13

Page 14: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

negative response. Measuring racial context at a smaller level would begin to mix simple

racial proximity with inter-personal interaction and self-selection into integrated neighbor-

hoods. Our second reason is pragmatic – the county-level is the smallest unit of geography

available to us in both data sets we employ in this study.11

This measure of racial context ranges from 0.1 to 72.3 with a mean of 10.68 and a standard

deviation of 11.75 among CCES respondents. It ranges 0.5 to 66.3 with a mean of 11.51 and

standard deviation of 12.09 among ANES respondents. As outlined above, the racial threat

hypothesis predicts that white voters will be less likely to vote for Obama and would be

more likely to have negative views of him as the size of the black population in their county

increases. Of course, we noted above that several scholars have reported positive rather than

negative effects, and still others have reported mixed findings.

Dependent Variables

We examine several dependent variables to test the effect of racial prejudice and racial

context on whites’ behaviors and attitudes during the 2008 Presidential election. One de-

pendent variable is the presidential vote choice reported by respondents. We measure this

variable as a dummy variable coded 1 for respondents who reported having voted for Obama

and 0 for those reporting having voted for McCain. Just over 45% of the white respondents

in our CCES sample and about 44% of white ANES respondents reported voting for Obama.

We also examine several measures of how respondents evaluated the two major party

nominees. The CCES asked respondents whether or not they thought each candidates was

“honest,” “knowledgeable,” and “experienced.” For each trait, we code the variable equal

to 1 if the respondent feels the candidate has the trait in question and 0 if they do not.

11Respondents’ zip codes are available for CCES respondents, but not for ANES respondents. The sameanalyses reported below was conducted at the zip code level with the results — direction of coefficients andlevels of significance — being unaffected.

14

Page 15: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

We use these measures to construct an additive measure of overall candidate evaluation for

both Barack Obama and John McCain on a scale of zero to three. If a respondent felt

Obama had all three of these qualities, she would be given a three for this measure. For

these measures, higher values indicate high evaluations of the candidate in question. We

then analyze each trait for each candidate separately using logistic regression. This allows

us to consider a more precise analysis of how prejudice and/or racial context affected the

way in which whites viewed both Obama and McCain.12

Control Variables

Our models also include a number of control variables. First, we control for partisanship

using two dichotomous variables indicating whether or not a respondent is a Republican or a

Democrat, respectively. We also use the respondent’s self-reported ideology on the standard

seven point scale where the lowest value indicates that the respondent identifies as being

extremely liberal while the highest value indicates that the respondent identifies as being

extremely conservative.

We include a control variable for sex coded one for women and zero for men. Categorical

measures of individual-level education and income are also included in our models. Education

levels range from one to six, with one indicating no high school diploma to six indicating

that the respondent has a post graduate degree for CCES respondents. The median level of

education is “some college.” Education among ANES respondents is measured more directly

as the number of formal years of education experienced by a respondent. The mean number

of years of education for white ANES respondents was approximately 13.7. Income varies

on a 15 point scale on which one indicates that the respondent reported earning “less than

10,000 dollars” and 15 indicates that they reported earning “greater than 150,000 dollars” in

12The ANES lacked comparable measures, so our analysis of citizens’ perceptions of the candidates’ traitsis limited to the CCES data.

15

Page 16: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

the CCES data. The median level of income in these data “50,000 to 60,000 dollars.” In the

ANES data, income is measured on a 25 point scale with a mean of just under 17. Finally,

we control for born again Christian status as a religious preference. Individuals who classify

themselves as “born again” are coded one while those who do not are coded zero.

Analysis and Results

As a first step, we examine how racial prejudice and racial context might be related to each

other among white voters. We argue it is possible that racial stereotypes might be driven in

part by the racial context in which respondents live, as illustrated by Equation 2. We regress

each of our three measures of racial prejudice on the percent black in a county, controlling for

party identification, ideology, income, gender, education, status as a born again Christian,

and median county income. The results of these models are presented in Table 1. The

results show that there is not a statistically significant (p ≤ .05) relationship between the

racial context in which a white respondent lives and her level of racial prejudice among

CCES respondents. The coefficient for the ANES model appears to be differ significantly

from zero, though the cumulative probabilities plotted in Figure 1 suggest that this effect

is not substantively large. Overall, these findings suggest that the attitudes held by white

respondents in these two surveys regarding blacks as a group are not strongly driven by the

racial context in which they live.13

The results in Table 1 suggest that the racial stereotypes about blacks expressed by white

citizens are consistently related to their ideological dispositions across models and data sets.

Whites who are relatively more conservative tend to hold more negative views of blacks

compared to whites who are relatively more liberal. The attitudes of CCES respondents

13Including the race of the interviewer as a control variable in the models using ANES data does notalter the substance of our results. The CCES is administered online and thus is not subject to potentialinterviewer effects.

16

Page 17: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

were responsive to the level of education attained by whites. Not surprisingly, those with

higher levels of education tend to report more positive views of blacks as a group than do

those with lower levels of education. Racial stereotypes, then, appear to be in part a function

of political ideology and education.

[Table 1 about here.]

[Figure 1 about here.]

Next, we predict white respondents’ vote choices in a number of models, each of which

contains one of our measures of racial stereotypes.14 Tables 2 contains the results of two

models, both of which make use of either the four-point or the seven-point hardworking

racial stereotypes measure to examine the influence of racial prejudice and context on the

decision to vote for Barack Obama in the 2008 presidential election. These models allow

us to compare the results produced by similar measures using two different survey samples

taken during the 2008 election.

The results of both models reported in Table 2 yield similar findings — the degree to

which whites believe that blacks are hardworking is a powerful predictor of reported vote

choice. As respondents view blacks as being more hardworking, they are increasingly likely

to report having voted for Obama. The models differ on the effect of racial context on

white citizens’ vote choice; the results of the ANES model suggest that as the percentage of

county residents who are black increases, the likelihood that whites report voting for Obama

decreases. The estimated coefficient for county percent black in the CCES model, on the

other hand, fails to achieve a traditional level of significance and has a positive rather than

14Some might question whether a full analysis of both racial stereotypes and racial context should allowfor a possible interaction between the two. Such a model was illustrated above in Equation 4. We estimatedsuch models using racial context indicators taken at both the zip code and county level. None of theseanalyses produced substantively different results from those presented here. This suggests the effect of racialstereotypes on vote choice is not conditional on the environment in which citizens live.

17

Page 18: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

a negative sign. As expected, Republicans and those who identify as more conservative are

less likely to vote for Obama in 2008 while Democrats were more likely to report having

voted for him.

To evaluate the magnitude of the effects reported in Table 2, we plot the predicted

probabilities of voting for Obama drawn from both models across all values of the racial

stereotypes and racial context variables in Figure 2. Panels (a) and (b) represent the feeling

that blacks are hardworking. As evident in both the plot for the CCES and ANES, there is

a positive and significant relationship showing that as respondents report feeling that blacks

are more hardworking, their predicted probability of voting for Obama increases. In the

CCES plot, there are four categories on the feeling thermometer that correspond with a

predicted probability of voting for Obama ranging from .5 up to .8. In the ANES plot, the

hardworking scale has seven values ranging from a low predicted probability of voting for

Obama at .21 to a high of .6.

As evident in panel (c) of Figure 2, the predicted probability of voting for Obama remains

relatively constant across the entire spectrum of the percent black in each county among

CCES respondents. In the ANES plot in panel (d), however, there is a sharp decrease in the

predicted probability of voting for Obama as the percent black in the respondent’s county

increases. White respondents in counties with the lowest black population had a predicted

probability of voting for Obama of just under .6, while whites living in counties with the

highest black population had a predicted probability of voting for Obama of only .1. The

results produced by the ANES model support the traditional racial threat hypothesis.

[Table 2 about here.]

[Figure 2 about here.]

Table 3 contains the results of a model predicting vote choice for Obama using the CCES

data. Here we employ our more complex measure of racial stereotypes based on the factor

18

Page 19: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

score outlined above. As indicated by the estimated coefficients and their associated standard

errors, as individuals’ feelings about blacks as a group become more positive, the probability

of voting for Obama increases in a statistically significant manner (p ≤ .05). Much like

the previous models, party identification and ideology are also important predictors of vote

choice while racial context fails to achieve a significant effect on the likelihood of voting for

Obama.

Figure 3 plots the predicted probability of voting for Obama based on the coefficients

in Table 3. Panel (a) of Figure 3 shows that the predicted probability of voting for Obama

increases as attitudes about blacks become more positive. In this figure, moving from re-

spondents with the most negative views towards blacks to those with the most positive views

increased their predicted probability of voting for Obama by almost 30 percentage points.

This is certainly a large substantive effect comparing across individuals. However, another

way to evaluate the impact of stereotypes on the election is to access how many votes this

might have cost Obama on Election Day.

In order to evaluate the impact of whites holding negative racial stereotypes on the votes

they reported casting, we conducted two simulations. Each simulation is based on the results

of our logistic regression model presented in Table 3. We explored two hypothetical scenarios.

In the first scenario, we examined what the predicted voting behavior of white respondents

would be if all those who had a negative score on our racial stereotype measure actually had

a neutral score of zero. In the second scenario, we examined the predicted voting behavior

of white respondents if every respondent held the most positive view of blacks we observed

on our measure. We left all other values for each variable at their observed levels for the

respondents. We then generated predicted probabilities of voting for Obama under these two

scenarios using our estimates from Table 3 and compared them to the predicted probabilities

of voting for Obama based on the actual data.

19

Page 20: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Based on the actual data, the mean of the predicted probability of voting for Obama

is estimated to be 0.456. When we changed every negative score on the racial stereotype

measure that respondents held to a zero, the mean predicted probability of voting for Obama

increased by 1.4 percentage points. Looking more closely, our results reveal that 10 of the

455 respondents saw their predicted probability of voting for Obama switch from being

less than 0.5 to greater than 0.5 based on this simulation. When we re-ran the simulation

giving all respondents the highest observed value on our racial stereotype measure, the

average predicted probability of voting for Obama increased by 4.9 percentage points, and

26 respondents saw their predicted probability of voting for Obama switch from less than

0.5 to greater than 0.5.

At first glance, the results of these simulations suggest a much more muted impact

of racial stereotypes on the behavior of whites than the 30 percentage point swing our

cross-sectional analysis reported in Figure 3 suggests. That would be a misreading of these

results. Many respondents already had very high or very low probabilities of voting for

Obama based on other powerful predictors like party identification. That many of those

individuals became more or less supportive, but never switched sides, due to their views

regarding blacks as a group does not deny the power of racial stereotypes to affect such

individuals. Additionally, we would never expect to observe a 30 percentage point vote

swing in the aggregate based on any variable. The aggregation of individual preferences into

election outcomes concentrates attention on those voters nearest the .50/.50 dividing line.

Given that nearly 100 million white voters cast ballots in the 2008 Presidential election, our

simulations predict that anywhere from 2.2 to 5.7 million white voters would have shifted

from being just below the .50/.50 threshold of voting for Obama to just above it. We consider

this a substantial number.15

15We also conducted a third simulation in which we examined a hypothetical scenario in which all respon-dents who had a positive score on our racial stereotype measure instead had a neutral score. In other words,we wished to observe how many votes Obama would have lost due to the disappearance of positive racial

20

Page 21: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

In contrast, panel (b) of Figure 3 shows that the percent black in a white respondent’s

county has no meaningful effect on the predicted probability of voting for Obama. The

coefficient estimate, however is positive and the predicted probability line shows a slight

positive trend, which runs counter to Key’s (1949) racial threat hypothesis. This effect is

not, however, significantly different than a null effect.

[Table 3 about here.]

[Figure 3 about here.]

Next we examine the effects of racial stereotypes and racial context on the respondents’

evaluations of both presidential candidates in 2008. Table 4 reports the results of ordered

logistic models predicting respondents’ evaluations of both Obama and McCain. The depen-

dent variable is an additive score of three components of candidate quality – whether or not

the candidate is knowledgeable, honest, and experienced.16 The measure ranges from zero

to three, with higher scores indicating a higher perception of candidate quality.

As evidenced by the results reported in Table 4, the same variables that predicted vote

choice remain key predictors of how white respondents view Obama as a candidate. Racial

stereotypes are a significant (p ≤ .05) predictor of how whites view Obama as a candi-

date. We again find that the coefficient estimate operating on the percent black in the

county does not approach a traditional level of statistical significance. Much like our vote

choice model, party identification and ideology are significant predictors of evaluations of

attitudes among white citizens. The results of our simulation shows that the predicted probability of votingfor Obama switched from greater than 0.5 to less than 0.5 for 6 of 455 respondents. Our simulation thuspredicts that neutralizing white citizens’ positive racial attitudes would lead to Obama receiving approxi-mately 1.3 million fewer votes. Thus, while Obama may have received votes from sympathetic whites whomight not have supported him otherwise, such support was more than off-set by the loss of votes amongthose holding negative views toward blacks as a group.

16We report models estimating respondents’ views on whether or not Obama and McCain possessed eachof these traits in the appendix. The models suggest that the relationships we observe in Table 4 are notdriven primarily by any single component of our index.

21

Page 22: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Obama; Democrats feel substantially more positive towards Obama compared to indepen-

dents, though Republicans were not significantly more negative of their evaluations of Obama

than were independents. Whites reported relatively more negative evaluations of Obama as

they became more ideologically conservative.

Evaluations of McCain, on the other hand, appear to be responsive primarily to party

identification, ideology, and born again Christian identification. Republicans, born again

Christians, and conservatives all evaluated McCain more positively. Unlike evaluations of

Obama, evaluations of McCain do not appear to be affected by racial stereotypes or racial

context. Given that McCain was a white candidate, it may not be surprising that racial

attitudes and context did not affect evaluations of him, but demonstrating that this was the

case is important because it indicates that citizens use different information and attitudes to

form attitudes about black and white candidates, even when those black and white candidates

are running against each other.17

[Table 4 about here.]

[Figure 4 about here.]

We plot the cumulative probabilities of expressing each level of affect towards both Obama

and McCain across the values of our racial stereotypes measure in Figure 4. The results in

Figure 4 show that the more positive a white respondent’s views towards blacks was, the

warmer she felt towards Obama. The cumulative probabilities plotted in Panel (a) of Figure

4 suggest that those who held the most negative views of blacks were likely to express lower

17We also ran models using the ANES data. In these models, the dependent variables were affect felttowards each of the candidates as measured by feeling thermometers and the primary independent variablesof interest were our simple measure of racial stereotypes — the degree to which respondents agreed with thestatement that blacks are hardworking — and racial context at the county level. These models producedsubstantively identical results in the case of affect felt towards McCain. This additional model estimat-ing affect towards Obama produced similar results to those reported in Table 4 for our racial stereotypesindicator, but the county level percent black variable produced an estimated coefficient that was negativeand statistically significantly (p ≤ .05) different than zero, indicating that white respondents who lived incounties with higher black populations reported liking Obama less.

22

Page 23: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

levels of affect toward Obama while those who held more positive views towards blacks

were more likely to express higher levels of affect towards Obama. The flat cumulative

probabilities plotted in Panel (b) suggest that racial attitudes towards blacks do not predict

affect towards McCain very well when controlling for the other covariates in the model.

Conclusions

We have argued that a unified approach to analyzing the impact of individual-level prej-

udice and context on political attitudes is the most appropriate way to evaluate how white

voters respond to black political candidates. Previous scholarship focusing on one theoretical

motivation or another often misses or underestimates the complex role of race in electoral

politics. The evidence presented in our paper fits with decades of research that offers mixed

findings. In 2008, there is overwhelming evidence that racial attitudes were a driving influ-

ence in both vote choice and candidate evaluations. There is limited evidence, on the other

hand, that racial context influenced these factors.

We found strong and consistent evidence of a direct effect of racial prejudice on how

whites viewed Obama and their likelihood of voting for him. Even after controlling for

robust predictors such as party identification and self-placed ideology, whites who held more

negative views of blacks as a group were significantly less likely to view Obama as honest,

knowledgeable, or experienced. Such voters were also significantly less likely to report having

voted for Obama. Interestingly, these same prejudicial feelings did not inform how voters

evaluated Obama’s white opponent, John McCain. Negative views of blacks as a group

transferred directly to evaluations of the black candidate, but did not transfer beyond that

to his opponent. While Obama’s evaluations appear to have suffered due to white prejudice,

McCain’s evaluations did not appear to have been affected by prejudicial attitudes held by

white citizens.

23

Page 24: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

In contrast, we show mixed evidence that race had any contextual effect on white voters

in this election. Our measure of racial context, the percent black in each respondent’s

county, produced a statistically significant coefficient estimate only in the model estimating

vote choice using the ANES data. The coefficient operating on our contextual measure

never differs significantly from zero in any of the models using the CCES data. There is

a weak tendency toward positive coefficients for this variable in the CCES as evidenced by

the models predicting evaluations of Obama. However, even if there is a weak direct effect

that is positive, this weak pattern is offset by the weak negative effect racial density had on

holding negative racial stereotypes.

Our fundamental conclusion is not a surprising one — race matters. The exact mechanism

regarding how race matters, and to what degree, remains uncertain and conditional. In 2008,

implicit racial cues could very well have triggered racial prejudice in white voters. Nobody

had to cue the public that Obama was black. The South Carolina primary and some of the

language surrounding the “birther” movement (which argued Obama was born in Kenya)

may have also raised the salience of racial attitudes as well. Still, the mere presence of

a black candidate could very easily cue uncertainty based on Obama’s perceived policy

preferences. As Kinder and Sears (1981, 415) argue, “the magnitude of racial threat to an

individual white should be the product of affect about some end state and expectancy that it

may materialize...” Having no previous major party black presidential candidates may have

induced uncertainty about the manner in which Obama would govern and prioritize various

groups of citizens.

Although we have little evidence that demographic contextual effects influence candidate

evaluations and vote choice in the 2008 presidential election, this is not to say they should be

dismissed. Most of the evidence of social-learning and positive responses to black candidates

occur in metropolitan elections (e.g. Liu 2003; Carsey 1995). The negative findings and

null findings in terms of contextual influences are almost all tied to elections at the state or

24

Page 25: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

national level (Wright 1977; Huckfeldt and Kohfeld 1989; Carsey and Windett 2013, e.g.).

These mixed results may also be due to the unit of geographic analysis–with the metropolitian

election analyses relying on census block and tract data, while our analysis and others use

more broad measures of geographic proximity. Future studies should aim for a more nuanced

analysis

The electoral victory of Barack Obama marks another milestone in the history of racial

politics in the U.S. Many commentators speculated about the impact of race on this election,

and some have concluded that Obama’s victory is the death knell for racialized politics in

the U.S. Our analysis clearly suggests otherwise, as individual racial prejudice played an

important role in shaping the attitudes and behaviors of white voters. We have shown that

negative racial attitudes among white voters in 2008 may have reduced Obama’s vote total

by 2.2 to 5.7 million votes. The group of voters who did not vote for Obama due to his race

were not numerous enough to alter the outcome of the 2008 presidential election, but our

results imply that in more competitive contests, racial attitudes and their effects on both

public opinion and voting behavior may become a more important consideration for potential

candidates to take into account when deciding whether or not to enter a race and, if they

do, what kind of strategy to pursue. Continued research in this area should strive to further

explore the root causes of this racialized behavior, all the while noting the challenging nature

of this intricate process.

25

Page 26: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Appendix

Racial Stereotype Question Wording: CCES

In this appendix, we provide greater detail on our measures of racial stereotypes. [Note

to Editor/Reviewers: if the paper is accepted for publication, we are happy to post this

appendix online or include it as part of the published article.] The following is the wording

for the racial stereotype questions in the 2008 CCES. Note that the coding scheme was

reversed for our analyses.

UNC40 should always be listed first. UNC41, UNC42, and UNC43 should be randomized:

“Below are several characteristics often used to describe groups of people. Of course no

characteristic applies to all people in a group, but thinking in general terms, do you agree

or disagree that each of the following characteristics applies to the majority of African-

Americans?”

Columns:

1 Agree Strongly

2 Agree Somewhat

3 Disagree Somewhat

4 Disagree Strongly

Rows:

UNC40 TRUSTWORTHY FIXED

UNC41 VIOLENT

UNC42 COMPLAINERS

UNC43 HARDWORKING

26

Page 27: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Hardworking Stereotype Question Wording: ANES

Now I have some questions about different groups in our society. I’m going to show you

a seven-point scale on which the characteristics of the people in a group can be rated. In the

first statement score of one means that you think almost all of the people in that group tend

to be “hard-working.” A score of seven means that you think most people in the group are

“lazy.” A score of four means that you think that most people in the group are not closer

to one end or the other, and of course, you may choose any number in between.

Factor Analysis to Compute CCES Racial Stereotype Measure

Below are the results of an exploratory factor analysis. This shows a single significant

factor is returned from our 4 stereotype questions. We have a theoretical motivation that

these are all indicators of racial stereotypes. In the factor score construction, a single sig-

nificant factor is the equivalent of forcing a single factor in a confirmatory factor analysis

approach. We have done both and the results in our models remain unchanged.

[Table 5 about here.]

[Table 6 about here.]

Analysis of the Components of our Additive Measure of Affect

Table 7 reports the results of estimating separate logistic regression models for each

component of our additive measures of candidate evaluation. These results indicate that

racial stereotypes are an important predictors of perceptions of Obama’s honesty, knowledge,

and experience. In each case, more positive evaluations of blacks as a group on average lead

to a higher likelihood of attributing each trait to Obama. Like our models predicting vote

27

Page 28: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

choice using the CCES data set, we again observe no statistically significant effect of racial

context on how whites viewed Obama as a candidate. In each case, the estimated coefficients

are indistinguishable from zero.

[Table 7 about here.]

Our results differ for respondents’ evaluations of McCain. In no instance does our in-

dicator of racial attitudes towards blacks influence white citizens’ evaluations of McCain’s

qualities as a candidate. Thus there does not appear to be any evidence that whites formed

their opinions of McCain in response to their attitudes toward blacks as a group even though

McCain was running against a black opponent. If Obama was hurt by negative stereotypes

held by some white voters, McCain did not benefit directly in terms of garnering a more

positive evaluation from such voters. At best, he only benefited indirectly because voters

who held negative attitudes towards blacks were less likely to vote for Obama ceteris paribus

and were thus more likely to vote for McCain. Like Obama, evaluations of McCain appear to

also be largely independent of the racial context within which white voters find themselves.

The estimated effect was positive in all three models, but not significantly different from

zero in any of them.

We plot predicted probabilities for attributing each of the three traits for both candidates

across all values of the racial stereotypes measure in Figure 5. The top row of plots were pro-

duced from the models estimating evaluations of Obama and the bottom row were produced

by the models predicting evaluations of McCain. Each plot shows the predicted probability

that each candidate is viewed as being honest, knowledgeable, or experienced. As evident in

the top row, the predicted probability that respondents attribute each of the three traits to

Obama increases as racial attitudes towards blacks become more favorable. The evaluations

of Obama’s honesty and experience show the sharpest increases; the predicted probability of

respondents viewing Obama as holding these traits increases more than 40 percentage points

28

Page 29: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

from the least to the most favorable views of blacks.

[Figure 5 about here.]

The bottom row of plots in Figure 5 provide further evidence that attitudes towards

blacks do not affect the attribution of these qualities to McCain by white respondents. Each

of the predicted probability plots is essentially flat across all values of the racial stereotypes

factor score.

29

Page 30: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

References

Allport, Gordon. 1954. The Nature of Prejudice. Menlo Park: Addison-Wesley.

Ansolabehere, Stephen D. and Charles Stewart. 2009. “Amazing Race.” Boston Review 1:1.

Bell, Derrick. 1997. Faces at the Bottom of the Well: The Permanence of Racism. New

York: Basic Books.

Black, Earl and Merle Black. 1987. Politics and Socieity in the South. Cambridge, MA:

Harvard University Press.

Blalock, Hubert M. 1967. Toward a Theory of Minority-group Relations. New York: Wiley.

Block Jr., Ray. 2011. “Backing Barack Because He’s Black: Racially Motivated Voting in

the 2008 Election.” Social Science Quarterly 92:423–446.

Campbell, AL, C. Wong and J. Citrin. 2006. “”Racial threat”, Partisan Climate, and Di-

rect Democracy: Contextual Effects in Three California Initiatives.” Political Behavior

28(2):129–150.

Carsey, Thomas M. 1995. “The Contextual Effects of Race on White Voter Behavior: The

1989 New York City Mayoral Election.” Journal of Politics 57:2:221–228.

Carsey, Thomas M. and Jason H. Windett. 2013. “The Contextual Effects of Race, Racial

Representation, and Elite Campaign Cues on Voter Behavior in Statewide Races.” Na-

tional Political Science Review p. forthcoming.

Citrin, Jack, Donald Philip Green and David O. Sears. 1990. “White Reactions to Black

Candidates: When Does Race Matter?” Public Opinion Quarterly 54(1):74–96.

Giles, Michael and Melanie Buckner. 1993. “David Duke and Black Threat: An Old Hy-

pothesis Revisited.” Journal of Politics 53:3:702–713.

30

Page 31: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Giles, Michael and Melanie Buckner. 1996. “Beyond Racial Threat: Failure of an Old

Hypothesis in the New South: Comment.” Journal of Politics 58:4:1171–1180.

Glaser, J..M. 2003. “Social Context and inter-group Political Attitudes: Experiments in

Group Conflict Theory.” British Journal of Political Science 33:607–620.

Hajnal, Zoltan. 2001. “Whte Residents, Black Incumbents, and the Declining Racial Divide.”

American Political Science Review 95(3):603–617.

Hopkins, Daniel J. 2010. “Politicized Places: Explaing Where and When Immigrants Provoke

Local Opposition.” American Political Science Review 104(1):40–60.

Huckfeldt, Robert and Carol Weitzel Kohfeld. 1989. Race and the Decline of Class in Amer-

ican Politics. Urbana, IL: University of Illinois Press.

Hutchings, Vincent L., Hanes Walton Jr. and Andrea Benjamin. 2010. “The Impact of

Explicit Racial Cues on Gender Differences in Support for Confederate Symbols and Par-

tisanship.” The Journal of Politics 72(4):1–14.

Key, V.O. Jr. 1949. Southern Politics in State and Nation. New York: A. Knopf.

Kinder, Donald R. and D. Roderick Kiewiet. 1981. “Sociotropic Politics:The American

Case.” British Journal of Political Science 11:129–162.

Kinder, Donald R. and David O. Sears. 1981. “Prejudice and Politics: Symbolic Racism

Versus Racial Threats to the Good Life.” Journal of Personality and Social Psychology

40:414–431.

Kinder, Donald R. and Tali Mendelberg. 1995. “Cracks in American Apartheid: The Political

Impact of Prejudice among Desegregated Whites.” Journal of Politics 57(2):402–424.

31

Page 32: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Levine, Jeffrey, Edward G. Carmines and Paul M. Sniderman. 1999. “The Empirical Dimen-

sionality of Racial Stereotypes.” Public Opinion Quarterly 63(3):371–384.

Lewis-Beck, Michael S., Charles Tien and Richard Nadeau. 2010. “Obama’s Missed Land-

slide: A Racial Cost?” PS: Political Science January:69–76.

Liu, B. D. 2003. “Deracialization and Urban Racial Context.” Urban Affairs Review

38(4):572–591.

Lupia, Arthur and Mathew D. McCubbins. 1998. The Democratic Dilemma: Can Citizens

Learn What they Need to Know? Cambridge, U.K.; New York: Cambridge University

Press.

Mas, Alexandre and Enrico Moretti. 2009. “Racial Bias in the 2008 Presidential Election.”

American Economic Review 999:999.

McDermott, Monika. 1998. “Race and Gender Cues in Low-Information Elections.” Political

Research Quarterly 51(4):895–918.

Mendelberg, Tali. 2001. The Race Card: Campaign Strategy, Implicit Messages, and the

Norm of Equality. Princeton: Princeton University Press.

Oliver, J. Eric and Tali Mendelberg. 2000. “Reconsidering the Environmental Determinants

of White Racial Attitudes.” American Journal of Political Science 44:574–589.

Parker, Christopher, Mark Sawyer and Christopher Towler. 2009. “A Black Man in the

White House?” Du Bois Review 6:193–217.

Piston, Spencer. 2010. “How Explicit Racial Prejudice Hurt Obama in the 2008 Election.”

Political Behavior 32:431–451.

32

Page 33: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Schaffner, Brian. 2011. “Racial Salience and the Obama Vote.” Political Psychology

32(6):963–988.

Sears, David O., Colette Van Laar, Mary Carrillo and Rick Kosterman. 1997. “Is It Really

Racism?: The Origins of White Americans’ Opposition to Race-Targeted Policies.” Public

Opinion Quarterly 61(1):16–53.

Sniderman, Paul M. and Edward G. Carmines. 1997. Reaching Beyond Race. Harvard

University Press.

Tarman, Christopher and David O. Sears. 2005. “The Conceptualization and Measurement

of Symbolic Racism.” Journal of Politics 67(3):731–761.

Terkildsen, Nayda. 1993. “When White Voters Evaluate Black Candidates: The Processing

Implications of Candidate Skin Color, Prejudice, and Self-Monitoring.” American Journal

of Political Science 37(4):1032–1053.

Tesler, Michael. and David Sears. 2010. Obama’s Race: The 2008 Election and the Dream

of a Post-Racial America. University of Chicago Press.

Voss, D. S. and D. Lublin. 2001. “Black Incumbents, White Districts - An Appraisal of the

1996 Congressional Elections.” American Politics Research 29(2):141–182.

Voss, Stephen. 1996. “Beyond Racial Threat: Failure of an Old Hypothesis in the New

South.” Journal of Politics 58:4:1156–1170.

Williams, Linda. 1990. “White/Black Perceptions of the Electability of Black Political Can-

didates.” National Black Political Science Review 2:45–64.

Wright, Gerald C. 1977. “Contextual Models of Electoral Behavior: The Southern Wallace

Vote.” American Political Science Review 71:497–508.

33

Page 34: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Table 1: Racial Stereotypes as a Function of Racial Context

ANES CCES CCEShardworking hardworking stereotypes

County percent black -0.017* -0.005 -0.003(0.01) (0.01) (0.00)

Democrat -0.402 0.056 -0.083(0.36) (0.24) (0.10)

Republican -0.346 0.253 0.082(0.35) (0.23) (0.10)

Ideology: 7 point -0.145* -0.294* -0.126*(0.08) (0.07) (0.03)

County median income 0.000 -0.000* -0.000*(0.00) (0.00) (0.00)

Born again 0.417* -0.256 -0.055(0.19) (0.20) (0.09)

Female 0.003 0.075 0.025(0.18) (0.17) (0.07)

Level of education 0.025 0.216* 0.076*(0.05) (0.06) (0.03)

Age in years 0.007 0.008 0.003(0.01) (0.01) (0.00)

Income -0.028 -0.017 0.008(0.02) (0.03) (0.01)

Intercept 0.381(0.27)

Cut 1 -4.526* -3.925*(0.90) (0.66)

Cut 2 -2.868* -2.007*(0.85) (0.64)

Cut 3 -1.586* 0.133(0.85) (0.63)

Cut 4 0.208(0.85)

Cut 5 1.326(0.84)

Cut 6 2.520*(0.87)

BIC 1,417.397 1,203.106 1,196.912N 415 488 488

Note: Estimated ordered logistic regression coefficients are reportedfor the first and second models while OLS coefficients are reportedfor the third model. Standard errors are reported in parentheses.* = p ≤ .05 (one tailed)

34

Page 35: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Table 2: Vote Choice in the 2008 Presiden-tial Election: The Effect of Beliefs About theWork Ethics of Blacks

ANES CCESBlacks are hardworking 0.272* 0.538*

(0.15) (0.23)County percent black -0.035* 0.013

(0.02) (0.02)Democrat 1.882* 0.711*

(0.59) (0.44)Republican -1.282* -1.574*

(0.61) (0.49)Ideology: 7 point -0.555* -1.640*

(0.17) (0.21)County median income -0.000 0.000

(0.00) (0.00)Born again -0.719* -0.697*

(0.39) (0.40)Female -0.442 -0.644*

(0.39) (0.38)Level of education -0.119 0.045

(0.10) (0.14)Age in years -0.013 -0.006

(0.01) (0.01)Income 0.003 0.127*

(0.05) (0.06)Intercept 4.096* 4.951*

(1.85) (1.50)BIC 267.962 290.237N 324 455

Note: Estimated logistic regression coeffi-cients are reported. Standard errors are shownin parentheses.* = p ≤ .05 (one tailed)

35

Page 36: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Table 3: Racial Stereotypes and Vote Choice inthe 2008 Presidential Election

Estimated coefficientRacial stereotypes 0.503*

(0.24)County percent black 0.014

(0.02)Democrat 0.756*

(0.44)Republican -1.572*

(0.49)Ideology: 7 point -1.614*

(0.21)County median income 0.000

(0.00)Born again -0.726*

(0.40)Female -0.627*

(0.37)Level of education 0.054

(0.14)Age in years -0.005

(0.01)Income 0.120*

(0.06)Intercept 6.405*

(1.39)

BIC 291.456N 455

Note: Estimated logit coefficients are reported.Standard errors are shown in parentheses. Datawas drawn from the 2008 CCES.* = p ≤ .05 (one tailed)

36

Page 37: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Table 4: Additive Measure of Affect Felt To-wards Obama and McCain in 2008

Obama McCainRacial stereotypes 0.695* -0.065

(0.15) (0.15)County percent black -0.007 0.012

(0.01) (0.01)Democrat 0.900* -0.309

(0.32) (0.27)Republican -0.239 0.855*

(0.29) (0.34)Ideology: 7 point -0.964* 0.581*

(0.11) (0.09)County median income 0.000 0.000

(0.00) (0.00)Born again -0.359 0.659*

(0.25) (0.29)Female -0.320 0.381*

(0.23) (0.23)Level of education 0.099 0.007

(0.08) (0.08)Age in years -0.006 0.003

(0.01) (0.01)Income 0.036 -0.019

(0.04) (0.04)Cut 1 -5.181* -1.482*

(0.92) (0.82)Cut 2 -3.869* 0.816

(0.90) (0.78)Cut 3 -3.257* 2.294*

(0.88) (0.79)BIC 768.798 777.574N 428 417

Note: Estimated coefficients are reported.Standard errors are shown in parentheses.Data was drawn from the 2008 CCES.* = p ≤ .05 (one tailed)

37

Page 38: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Table 5: Factor Loadings For Racial StereotypeMeasure

Variable Factor1 Factor2 UniquenessTrustworthy 0.7939 -0.1957 0.3315Hard workers 0.7733 -0.2169 0.3549Complainers 0.7201 0.2331 0.4271Violent 0.7591 0.2045 0.3819

38

Page 39: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Table 6: Eigenvalues For CCES Racial Stereotype Factor Anal-ysis

Factor Eigenvalue Difference Proportion CumulativeFactor 1 2.32314 2.14165 1.0657 1.0657Factor 2 0.18149 0.33429 0.0833 1.1489Factor 3 -0.15280 0.01902 -0.0701 1.0788Factor 4 -0.17182 . -0.0788 1.0000

39

Page 40: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Tab

le7:

Can

did

ate

Eva

luat

ions

Can

did

ate

ishon

est

Can

did

ate

isknow

ledge

able

Can

did

ate

isex

per

ience

dO

bam

aM

cCai

nO

bam

aM

cCai

nO

bam

aM

cCai

nR

acia

lst

ereo

typ

es0.

706*

0.05

90.

466*

-0.0

010.

650*

0.34

1(0

.22)

(0.1

6)(0

.16)

(0.1

7)(0

.20)

(0.2

9)C

ounty

per

cent

bla

ck0.

002

0.00

4-0

.012

0.00

40.

000

0.00

2(0

.01)

(0.0

1)(0

.01)

(0.0

1)(0

.01)

(0.0

2)D

emocr

at0.

340

-0.3

470.

599

-0.3

560.

975*

0.72

2(0

.39)

(0.3

1)(0

.38)

(0.3

0)(0

.34)

(0.5

0)R

epublica

n-0

.768

*0.

700*

-0.0

410.

439

-1.1

70*

0.07

3(0

.42)

(0.3

4)(0

.30)

(0.4

1)(0

.42)

(0.7

2)Id

eolo

gy:

7p

oint

-1.4

40*

0.67

4*-0

.814

*0.

467*

-0.8

93*

0.47

4*(0

.17)

(0.1

1)(0

.12)

(0.1

1)(0

.13)

(0.1

9)C

ounty

med

ian

inco

me

0.00

0-0

.000

0.00

0-0

.000

0.00

00.

000

(0.0

0)(0

.00)

(0.0

0)(0

.00)

(0.0

0)(0

.00)

Bor

nag

ain

-0.2

250.

470

-0.4

130.

399

-0.3

232.

126*

(0.3

6)(0

.30)

(0.2

7)(0

.34)

(0.3

3)(1

.06)

Fem

ale

-0.2

660.

314

0.08

50.

344

-0.5

69*

-0.9

94*

(0.3

2)(0

.25)

(0.2

5)(0

.26)

(0.3

0)(0

.46)

Lev

elof

educa

tion

0.07

00.

044

0.14

8-0

.010

0.14

60.

073

(0.1

2)(0

.09)

(0.0

9)(0

.09)

(0.1

1)(0

.16)

Age

inye

ars

-0.0

200.

014

-0.0

10-0

.006

-0.0

06-0

.006

(0.0

1)(0

.01)

(0.0

1)(0

.01)

(0.0

1)(0

.02)

Inco

me

0.03

50.

028

0.00

1-0

.041

0.02

2-0

.029

(0.0

6)(0

.04)

(0.0

4)(0

.04)

(0.0

5)(0

.08)

Inte

rcep

t6.

946*

-3.4

11*

4.32

9*0.

256

2.83

3*0.

608

(1.3

2)(0

.91)

(1.0

1)(0

.92)

(1.0

7)(1

.56)

BIC

340.

362

495.

103

483.

097

471.

113

399.

212

236.

353

N45

444

946

245

346

647

3

Not

e:E

stim

ated

logi

tco

effici

ents

are

rep

orte

d.

Sta

ndar

der

rors

are

show

nin

par

enth

eses

.D

ata

isdra

wn

from

the

2008

CC

ES.

*=

p≤

.05

(one

tailed

)

40

Page 41: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

County Percent Black

Cum

ulat

ive

Pro

babi

lity

of T

hink

ing

Bla

cks

are

Incr

easi

ngly

Har

dork

ing

0 10 20 30 40 50 60

00.

20.

40.

60.

81

Note: Cumulative probabilities were generated from the results reported in the first column of Table 1. These

are the cumulative probabilities of evaluating blacks as being various levels of hardworking across all values

of county percent black for white male independents who did not report that they were born again and who

identified as ideological moderates. The rest of the variables in the models were held at their means. From

top to bottom, each line represents an increasingly high evaluation.

Figure 1: Cumulative Probabilities of Believing Blacks are Hardworking

41

Page 42: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Bla

cks

are

Har

dwor

king

Predicted Probability of Voting for Obama

12

34

00.20.40.60.81

(a)

CC

ES

har

dw

ork

ing

Bla

cks

are

Har

dwor

king

Predicted Probability of Voting for Obama

12

34

56

7

00.20.40.60.81

(b)

AN

ES

hard

work

ing

Cou

nty

Per

cent

Bla

ck

Predicted Probability of Voting for Obama

010

2030

4050

6070

00.20.40.60.81

(c)

CC

ES

raci

alco

nte

xt

Cou

nty

Per

cent

Bla

ck

Predicted Probability of Voting for Obama

010

2030

4050

60

00.20.40.60.81

(d)

AN

ES

raci

al

conte

xt

Note

:P

redic

ted

pro

babi

liti

esw

ere

gen

erate

dfr

om

the

resu

lts

repo

rted

inT

abl

e2.

Thes

eare

the

pre

dic

ted

pro

babi

liti

esof

voti

ng

for

Oba

ma

acr

oss

all

valu

esof

the

dep

enden

tva

riabl

esof

inte

rest

for

whit

em

ale

indep

enden

tsw

ho

did

not

repo

rtth

at

they

wer

ebo

rnaga

inan

dw

ho

iden

tifi

edas

ideo

logi

cal

mod

era

tes.

The

rest

of

the

vari

abl

esin

the

mod

els

wer

ehel

dat

thei

rm

ean

s.

Fig

ure

2:W

hit

es’

Vie

ws

onB

lack

s’W

ork

Eth

ics,

Rac

ial

Con

text,

and

Vot

ing

for

Obam

a

42

Page 43: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Rac

ial S

tere

otyp

es

Predicted Probability of Voting for Obama

−2

−1.

5−

1−

0.5

00.

51

00.20.40.60.81

(a)

Rac

ial

ster

eoty

pes

Cou

nty

Per

cent

Bla

ck

Predicted Probability of Voting for Obama

010

2030

4050

6070

00.20.40.60.81

(b)

Raci

al

conte

xt

Note

:P

redic

ted

pro

babi

liti

esw

ere

gen

erate

dfr

om

the

resu

lts

repo

rted

inT

abl

e3.

Thes

eare

the

pre

dic

ted

pro

babi

liti

esof

voti

ng

for

Oba

ma

acr

oss

all

valu

esof

the

raci

al

ster

eoty

pes

vari

abl

efo

rw

hit

em

ale

indep

enden

tsw

ho

did

not

repo

rtth

at

they

wer

ebo

rnaga

inan

dw

ho

iden

tifi

ed

as

ideo

logi

cal

mod

erate

s.T

he

rest

of

the

vari

abl

esin

the

mod

els

wer

ehel

dat

thei

rm

ean

s.

Fig

ure

3:R

acia

lSte

reot

yp

esan

dth

eP

redic

ted

Pro

bab

ilit

yof

Vot

ing

for

Obam

a

43

Page 44: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Rac

ial S

tere

otyp

es F

acto

r S

core

Cumulative Probability of Feeling Increasingly Positive Towards Obama

−2

−1

01

00.20.40.60.81

(a)

Ob

ama

Ad

dit

ive

Aff

ect

Sco

re

Rac

ial S

tere

otyp

es F

acto

r S

core

Cumulative Probability of Feeling Increasingly Positive Towards McCain

−2

−1

01

00.20.40.60.81

(b)

McC

ain

Ad

dit

ive

Aff

ect

Sco

re

Note

:C

um

ula

tive

pro

babi

liti

esw

ere

gen

erate

dfr

om

the

resu

lts

repo

rted

inT

abl

e4.

Thes

eare

the

cum

ula

tive

pro

babi

liti

esof

feel

ing

vari

ou

s

leve

lsof

aff

ect

tow

ard

sO

bam

aan

dM

cCain

acr

oss

all

valu

esof

the

raci

al

ster

eoty

pes

mea

sure

for

whit

em

ale

Rep

ubl

ican

sw

ho

did

no

tre

port

that

they

wer

ebo

rnaga

inan

dw

ho

iden

tifi

edas

ideo

logi

cal

mod

erate

s.T

he

rest

of

the

vari

abl

esin

the

mod

els

wer

ehel

dat

thei

rm

ean

s.F

rom

the

bott

om

of

each

figu

reto

the

top,

each

lin

ere

pre

sen

tth

ecu

mu

lati

vepro

babi

lity

that

indiv

idu

als

felt

at

least

that

leve

l(z

ero

thro

ugh

thre

e)of

aff

ect

tow

ard

sth

eca

ndid

ate

s.

Fig

ure

4:R

acia

lSte

reot

yp

esan

dA

ffec

tF

elt

Tow

ards

Pre

siden

tial

Can

did

ates

44

Page 45: Racial Stereotypes, Racial Context, and the 2008 ...kevinbanda.weebly.com/uploads/2/1/6/4/21649732/... · These works use the ANES questions on racial attitudes towards blacks in

Rac

ial S

tere

otyp

es

Predicted Probability of Believing Obama is Honest

−2

−1.

5−

1−

0.5

00.

51

00.20.40.60.81

Rac

ial S

tere

otyp

es

Predicted Probability of Believing Obama is Knowledgeable

−2

−1.

5−

1−

0.5

00.

51

00.20.40.60.81

Rac

ial S

tere

otyp

es

Predicted Probability of Believing Obama is Experienced

−2

−1.

5−

1−

0.5

00.

51

00.20.40.60.81

Rac

ial S

tere

otyp

es

Predicted Probability of Believing McCain is Honest

−2

−1.

5−

1−

0.5

00.

51

00.20.40.60.81

(a)

Hon

est

Rac

ial S

tere

otyp

es

Predicted Probability of Believing McCain is Knowledgeable

−2

−1.

5−

1−

0.5

00.

51

00.20.40.60.81

(b)

Kn

owle

dgea

ble

Rac

ial S

tere

otyp

es

Predicted Probability of Believing McCain is Experienced

−2

−1.

5−

1−

0.5

00.

51

00.20.40.60.81

(c)

Exp

erie

nce

d

Note

:P

redic

ted

pro

babi

liti

esw

ere

gen

erate

dfr

om

the

resu

lts

repo

rted

inT

abl

e7.

Thes

eare

the

pre

dic

ted

pro

babi

liti

esof

thin

kin

gth

at

Oba

ma

an

dM

cCain

are

hon

est,

know

ledge

abl

e,an

dex

peri

ence

dacr

oss

all

valu

esof

the

raci

al

ster

eoty

pes

mea

sure

for

whit

em

ale

Rep

ubl

ican

sw

ho

did

not

repo

rtth

at

they

wer

ebo

rnaga

inan

dw

ho

iden

tifi

edas

ideo

logi

cal

mod

erate

s.T

he

rest

of

the

vari

abl

esin

the

mod

els

wer

ehel

dat

thei

r

mea

ns.

The

top

row

of

plo

tsare

evalu

ati

on

sof

Oba

ma

whil

eth

eplo

tsin

the

bott

om

row

are

evalu

ati

on

sof

McC

ain

.

Fig

ure

5:P

redic

ted

Pro

bab

ilit

yof

Att

ributi

ng

Char

acte

rist

ics

toO

bam

aan

dM

cCai

n

45