1
Divided by Skin Color: The Influence of Skin Tone and Racialized News Messages on Latino Partisanship
Spencer Piston Syracuse University
Logan Strother
Syracuse University
Acknowledgments
We thank, first and foremost, the instructors of the ICPSR course “Designing, Conducting, and Analyzing Multi-Racial and Ethnic Political Surveys,” Lorrie Frasure-Stokely, Ricardo Ramirez, Gabe Sanchez, and Janelle Wong; course participants Chinbo Chong, Na Youn Lee, and Vanessa Cruz Nichols also provided valuable feedback. The experimental design was informed by suggestions from Scott Clifford, Thomas Leeper, Michele Margolis, Tim Ryan, and Emily Thorson. A previous version of this paper was presented at the 2015 Annual Meeting of the Midwest Political Science Association; we thank attendees for their feedback, especially Seth Goldman, the discussant, and Zoltan Hajnal. Participants at the Center for Political Studies Workshop at the University of Michigan also provided helpful comments, especially John Garcia, LaGina Gause, Vince Hutchings, Hakeem Jefferson, Don Kinder, Rob Mickey, Fabian Neuner, Rosemary Sarri, Chris Skovron, Lester Spence, and Nicole Yadon. Scholars at Stony Brook University also provided helpful comments during a presentation there, especially Alexa Bankert, Jason Barabas, Richard Cho, Andy Delton, Peter DeScioli, Leonie Huddy, Stanley Feldman, Reuben Kline, Yanna Krupnikov, Milton Lodge, Ben Newman, John Ryan, Jeff Segal, Yamil Velez, and Johanna Willmann. Finally, we thank Matt Barreto, Scott Clifford, Jenn Chudy, Nathan Kalmoe, Kristyn Karl, Adam Seth Levine, Ian Haney Lopez, Roque Planas, Karthick Ramakrishnan, Sean Richey, Tim Ryan, John Sides, Nadav Tanners, and Michael Tesler for their important reactions to early results from this research project, some of which were presented on The Monkey Cage Blog at www.washingtonpost.com.
2
Abstract
In light of the changing racial demographics of the new century, we extend scholars’
understanding of race and partisan preferences by examining two interacting factors shaping
Latino partisanship: phenotypic variation in skin tone and media-driven racialized images of the
parties. We argue that skin tone shapes the choice set available to Latinos as they navigate an
increasingly biracial (White/Black) partisan landscape in the United States. Examining three
nationally representative survey datasets, we find that light-skinned Latinos are less likely than
dark-skinned Latinos to identify as Democrats and to vote for Democrats, especially in recent
years. We also present evidence of the mechanism underlying this relationship, demonstrating
experimentally that news media communications linking the Republican Party with Whites and
the Democratic Party with Blacks strengthen associations between Latino skin tone and
partisanship. The concluding section discusses implications of these findings for the study of
partisanship and race in the twenty-first century.
3
Party identification has been described as “the key concept of U.S. electoral research”
(MacKuen, Erikson, and Stimson 1989), and it plays a central role in the formation of public
opinion as well (e.g., Slothuus and de Vreese 2010). Notably, recent research on mass
partisanship finds substantial evidence of partisan social polarization (Iyengar, Sood, and Lelkes
2012; Mason 2015). That is, partisans are increasingly active, biased, angry, and likely to dislike,
even loathe, their opponents. Relatedly, scholarship has also identified racial polarization of the
mass base of the two major political parties. For example, over the course of Obama’s first term
as president, white identification with the Democratic Party declined while black identification
with the Democratic Party increased (Tesler forthcoming). Furthermore, among whites, partisan
identification is increasingly bound up with racial attitudes (Sears and Valentino 2005),
including opposition to interracial marriage (Tesler 2013). Among Blacks, meanwhile, partisan
identification is also tied up with racial attitudes, namely group solidarity (Tesler forthcoming).
Vote choice in the past two presidential elections has also been influenced more by racial
prejudice than it was in previous elections (Kinder and Dale-Riddle 2012; Pasek et al. 2014;
Piston 2010), as has vote choice in Congressional elections (Luttig 2015).
Existing scholarship, then, tells us a lot about mass partisanship – about its determinants,
its importance for electoral outcomes and public opinion, and its social and racial polarization in
recent years. Yet extant research, and in particular research on the topic of race and partisanship,
also has important limitations. First, it focuses primarily on White and Black Americans,1 even
as racial demographics in the U.S. are rapidly undergoing dramatic changes. In particular,
Latinos, an admittedly broad classification that incorporates many different nationalities, have
recently eclipsed African Americans as the country’s largest single ethnic minority.
1 Tesler (forthcoming) focuses on the attitudes of Whites and Blacks but also includes valuable analyses of Latinos in Chapter 8.
4
Demographers predict that by 2050, non-Hispanic Whites will no longer constitute a majority of
the country’s population. Second, scholarship that does examine the determinants of partisanship
among Latinos (e.g., Hajnal and Lee 2011) has yet to explore the possibility that different
members of the broad category “Latino” are differentially racialized depending on an important
source of phenotypic heterogeneity – skin tone – with critical implications for Latino
macropartisanship. Third, existing research does not often consider the impact of political
communication that emphasizes the (changing) racial constituencies of the two major parties;
news media sources routinely perpetuate the image of the GOP, for example, as the party of
white people (Tesler forthcoming).
This project extends scholars’ understanding of race and partisanship in three ways: by
moving beyond the traditional focus on Whites and Blacks, examining the determinants of
partisanship among Latinos; drawing attention to an often-overlooked factor shaping Latino
partisan preferences, phenotypic variation in skin tone; and assessing the impact of changing
images of the two major parties as perpetuated by the news media. The first component of this
project develops theoretical expectations about the uneasy partisan status of Latinos in a “biracial
system” dominated by a black/white boundary (Bonilla-Silva and Dietrich 2009). Next, using
observational data from multiple national surveys across time, we demonstrate that skin color is
powerfully associated with Latino partisanship, such that light-skinned Latinos are less likely
than dark-skinned Latinos to identify as Democrats, even after controlling for standard factors
such as country of heritage and socioeconomic status. We also present the results of two original
experiments, allowing us to gain confidence in causal inferences about the impact of skin tone on
partisanship. We first show that priming skin tone through subtly manipulating its order in a
survey questionnaire is sufficient to increase its association with partisanship. Second, in a
5
parallel study, we examine the mechanism behind the skin tone/partisanship relationship among
Latinos. We manipulate whether subjects are exposed to standard news communication
perpetuating racialized party images. The findings reveal that when the news media link the
Republican Party to Whites and the Democratic Party to Blacks, this polarizes Latinos by skin
tone: light-skinned Latinos become more likely to identify as Republicans, and dark-skinned
Latinos become more likely to identify as Democrats.
Based on these findings, we argue in the concluding section that in order to understand
the complex racial dynamics of partisanship in American society, the research community must
move beyond a binary conception of U.S. racial politics. As Whites, especially Whites who are
prejudiced against Blacks, move to the Republican Party, and as Blacks, especially group-
conscious Blacks, solidify their relationship with the Democratic Party, this increasingly stark
racial partisan divide has important consequences for Latino partisanship. That is, as news media
promulgate the image of the GOP as the party of Whites, dark-skinned Latinos have become
more Democratic, and light-skinned Latinos have become less so. Even within the racialized
category “Latino,” partisans are increasingly “divided by color.”
Theoretical Expectations: The Relationship between Skin Tone and Latino Partisanship
The United States racial hierarchy has often been described as a “biracial system”
dominated by a White/Black boundary (Bonilla-Silva and Dietrich 2009). As Wimmers (2008)
observes, “the “one drop rule” draws a sharp line between “black” and “white” and is largely
accepted by individuals on both sides.” Latinos do not fit easily into this dichotomy, yet they are
not free to reshape it – indeed, they are often asked to situate themselves within it. For example,
6
federal policy often sets apart Hispanics and Latinos aside as an ethnic group and then asks them
to decide which racial category best describes them.
Patterns of macropartisanship in the United States reflect this Black/White paradigm.
Since the 1960s, Blacks have voted Democratic in overwhelming proportions, while Whites have
leaned Republican (Hutchings 2009; Kinder and Sanders 1996). Republicans are widely
perceived to be more likely than Democrats to represent White interests, while Democrats are
widely perceived to be more likely than Republicans to represent Black interests – a fact that
explains why racially prejudiced Southern whites have continued to migrate to the Republican
Party in recent decades (Valentino and Sears 2005). Latinos thus encounter not only a biracial
society in general but a biracial partisan landscape in particular.
Comparative studies of ethnicity document numerous instances of “strategic
repositioning” in which members of low-status ethnic groups attempt to expand the boundaries
of a dominant ethnic group in order to reap the benefits of that group’s status (Wimmer 2008). In
the case of the United States, given Whites’ dominant position in the racial hierarchy, it is in
Latinos’ interest to affiliate themselves with Whites when possible. Consistent with this
perspective, sociological research has routinely found that “most Latinos recognize the
advantages of a White racial designation when asked to self-identify” (Frank, Akresh, and Lu
2010).
But the opportunity to affiliate oneself with Whites is not uniformly present. In particular,
we argue that skin tone shapes the choice set available to Latinos as they encounter the racialized
political system in the United States. Existing research documents substantial porosity in the
White/Latino boundary, especially in comparison with the White/Black boundary. For example,
Whites are much more likely to marry Latinos and live in neighborhoods with Latinos than they
7
are to marry or live next to Blacks (Bonilla-Silva and Dietrich 2009). For those with the
opportunity of affiliating with Whites – of becoming “honorary Whites,” as Bonilla-Silva and
Dietrich (2009) put it, taking advantage of this opportunity comes with an increase in status. But
phenotype – in particular, skin color – determines whether this option exists. Whites may be
more likely to discriminate against dark-skinned Latinos than light-skinned Latinos in the labor
market (Frank, Akresh, and Lu 2010), and whites may be less willing to marry – or even live
next to – dark-skinned Latinos as well (Bonilla-Silva and Dietrich 1999). Light-skinned Latinos,
then, are more likely to have an option that dark-skinned Latinos do not: the option of affiliating
themselves with whites and distancing themselves from blacks.
The consequences of skin tone for Latino partisanship, therefore, are as follows. Research
indicates that the two major parties are evaluated in large part based on the social groups they are
thought to represent (e.g., Green, Palmquist, and Schickler 2004; Kinder and Dale-Riddle 2011).
To the extent that light-skinned Latinos have the opportunity to affiliate themselves with Whites
and distance themselves from Blacks, they should be less likely to align themselves with the
party perceived to represent Blacks, the Democratic Party, and more likely to align themselves
with the party perceived to represent Whites, the Republican Party.
Of course, skin tone is not expected to be the only factor influencing Latinos’
partisanship, or even the primary one. Other determinants include economic status (Hajnal and
Lee 2011), religion (Lee and Pachon 2007), country of heritage (Bonilla-Silva and Dietrich
2009), and acculturation (Cain, Kiewiet, and Uhlaner 1991). Given the importance of these
factors, we should not expect to find, for example, that the modal light-skinned Latino is a card-
carrying Republican. Yet we expect that skin tone nonetheless plays a powerful role in Latino
partisan preferences and voting decisions.
8
Empirical Analyses
Observational Analyses
We have argued that light skin tone leads some Latinos to affiliate themselves with
Whites, and accordingly with the party thought to represent Whites: the Republican Party. If this
is correct, we ought to witness a relationship between skin tone and partisanship among Latinos,
as well as a relationship between skin tone and vote choice. Furthermore, we should see that
among Latinos light skin tone is positively associated with favorable views of Whites and
unfavorable views of Blacks.
To subject these propositions to empirical scrutiny, we conduct analyses of nationally
representative survey data, focusing on the 2012 American National Election Studies (ANES)
time series survey. This survey suits our purposes admirably for a few reasons. First, it includes
an oversample of Latino citizens. Second, the quality of this sample is high: address-based
sampling was used and Spanish-language interviews were conducted in some cases in order to
ensure that the survey could reach a wider range of Latinos (n.b., Lee and Perez 2014). Third, the
dataset includes measures of the concepts central to the argument advanced here: skin tone,
partisanship, vote choice, and racial attitudes.
Only the face-to-face component of the 2012 ANES is analyzed, since the Internet
component did not include a measure of skin tone. Of the 472 Latinos in the face-to-face
component, 458 have non-missing values for skin tone, and 431 of these have non-missing
values for all control variables in the regression analyses; listwise deletion is used to deal with
missing data.
9
Skin color was measured by the interviewer at the beginning of the pre-election wave of
the 2012 survey. See Figure 1 for details; note that in the analyses presented below, the scale is
reversed such that higher values indicate lighter skin tone. Ethnicity and racial identification
were measured separately, and the racial identification question allowed respondents to select
more than one category, allowing individuals, for example, to identify simultaneously as Latino
and white (and/or black). Party identification was measured through a standard seven-point scale,
and racial attitudes were measured through standard stereotype batteries and group thermometer
scores (n.b., Hutchings 2009; Piston 2010). Exact question wording for measures of ethnicity,
racial identification, and racial attitudes can be found in Appendix 1.
[Figure 1 about here]
The analyses of the 2012 ANES are supplemented with analyses of two additional
nationally representative survey datasets: the 2008 Collaborative Multi-racial Post-Election
Survey and the 1989-1990 Latino National Politics Study. These studies also include large
samples of Latinos, including Spanish-language interviews and weights for national
representativeness. Because these studies measure skin tone differently, asking respondents to
rate their own skin color on a 1 to 5 scale, the inclusion of these datasets allows us to assess the
robustness of any results we find across measures of this key independent variable. Unless
otherwise indicated, all analyses of these survey datasets are weighted for national
representativeness. Throughout the manuscript, all statistical tests reported are one-tailed, and all
variables are transformed onto a 0 to 1 scale.
Experimental Studies
10
We also build on the observational studies with two experiments, which are designed to
provide analytical leverage on the question of causation. First, if skin tone truly shapes Latino
partisan preferences, we should find that priming skin tone increases its relationship with
partisanship. To investigate this possibility, we conducted an experiment on Latino respondents
from Amazon’s Mechanical Turk in August 2015 (MTurk: N = 806); while samples from MTurk
are not representative of the national population, many canonical experiments have been
replicated on MTurk survey participants (Berinsky, Huber, and Lenz 2013; Weinberg, Freese,
and McElhattan 2014). The experiment manipulates the order of the skin tone measure in the
survey questionnaire. Subjects are divided into two groups: in the Priming Condition,
respondents are asked to rate their skin tone (using the same measure as in the 2012 ANES) just
before they are asked their partisanship, while in the Control Condition, respondents are not
asked to rate their skin tone until after the partisanship question, at the end of the survey
questionnaire.
We have also hypothesized a mechanism for the Latino skin tone/partisan relationship:
that light-skinned Latinos seek to associate themselves with whites, leading some to affiliate with
the Republican Party. If this is correct, the skin tone/partisanship relationship should increase
when the Republican Party’s ties to whites and the Democratic Party’s ties to blacks are made
salient. To examine this proposition, an additional experiment, also conducted on MTurk in
August 2015 (N = 603),2 manipulated the salience of the perceived racial constituencies of the
two major parties. This manipulation took place as follows: subjects were randomly assigned to
three groups. One-third of the respondents, those in the Racialized News Condition, read a
2 In order to maximize statistical power, these two experiments were conducted simultaneously, and they share a control condition, in which skin tone was asked at the end of the questionnaire and subjects did not read about racialized constituencies of the parties.
11
(fictional) newspaper article – a modified version of an actual article3 about the racial
composition of Democratic and Republican voters in the 2012 election. Another third were
randomly assigned to the Non-Racialized News Condition, and these respondents were asked to
read an article that was nearly identical to the first except that mentions of race were stripped out.
The final third of the respondents, those in the Control Condition, were not asked to read
anything: these latter two conditions establish two different baselines against which to compare
the condition in which respondents were exposed to racialized descriptions of the two major
parties. Afterward, all respondents were asked the standard seven-point partisanship question,
enabling us to assess the impact of news media communication on the relationship between skin
tone and partisanship. Experimental stimuli can be found in Appendix 2.
Results
Distribution of Skin Tone
The distribution of skin tone among Latinos in the 2012 ANES can be found in Figure 2.
As the figure shows, the vast majority of Latinos fall on the light end of the skin tone scale
(reverse-coded from 0 to 10). Indeed, about 92% of the sample falls in the lightest six categories,
while the remaining 8% is distributed among the darkest five categories. This skewed
distribution affects the interpretation of regression coefficients, as will be addressed later; also
noteworthy is that substantial variation in skin tone is evident in the sample. Furthermore, such
variation is not solely a function of country of heritage. Additional analyses examining
3 Nate Cohn, “Southern Whites’ Loyalty to GOP Nearing That of Blacks to Democrats,” New York Times, April 23, 2014. Last accessed on October 22, 2015, at: http://www.nytimes.com/2014/04/24/upshot/southern-whites-loyalty-to-gop-nearing-that-of-blacks-to-democrats.html?_r=0
12
exclusively those Latinos who identify as Mexican, Mexican American, or Chicano (N = 305)
yield no meaningful change to the distribution.
[Figure 2 about here]
Assessing the Relationship Between Latino Skin Tone and Partisan Identification
If we are correct that skin tone shapes Latinos’ partisan preferences, we ought to see
associations between skin tone and party identification. To test this hypothesis, below we present
the results of three separate ordinary least squares regression models: one from the 1989-1990
LNPS, one from the 2008 CMPS, and one from the 2012 ANES. In each of these models, the
dependent variable is the standard seven-point party identification measure, ranging from Strong
Republican (low values) to Strong Democrat (high values).
In each of these regression models, the independent variable of interest is skin tone. The
models also control for a series of variables plausibly related to both skin color and partisan
preferences: (1) demographics: education, age, gender, income, and home ownership; (2) nativity
and language of interview; (3) country or region of heritage; and (4) religion. Demographic
variables are included given the relationship between skin color and socioeconomic status among
Latinos (Frank, Akresh, and Lu 2010) and the long-standing relationship between socioeconomic
status and party identification (e.g., Brooks and Manza 1997). Nativity and language of interview
are included as a result of the relationship between acculturation and partisanship (Cain, Kiewiet
and Uhlaner 1991) and a potential relationship between acculturation and skin tone, while
heritage is included due to the possibility that any relationship between skin tone and
partisanship actually reflects differences between, for example, between Cuban Americans and
Mexican Americans (n.b., Bonilla-Silva and Dietrich 2009). Finally, religion is included due to
13
its association with vote choice among Latinos (Lee and Pachon 2007) and the possibility that
Latinos with different skin tones might have different religious affiliations.
As Table 1 shows, skin tone is strongly associated with party identification across all
three datasets: Latinos with light skin are less likely to identify as Democrats. In the 2012 ANES,
this relationship is particularly strong; as the coefficient indicates, movement from the darkest to
the lightest skin tone values is associated with movement of nearly one-quarter across the party
identification scale. Furthermore, this relationship is statistically significant at the p<0.01 level.
To be sure, as mentioned above, the distribution of the skin tone variable is skewed toward the
lighter end. But even accounting for this by exclusively examining the results for just those 92%
of Latinos in the sample at the light end of the skin tone measure still yields a large association
of one-seventh of the scale, or one full point of the seven-point partisanship measure.
Strong as this relationship is, it is important to note its limits. For one thing, the
coefficient on skin tone and partisanship is substantially smaller in the 1989-1990 LNPS and the
2008 CMPS than in the 2012 ANES. This is consistent with the argument advanced here, which
is that the increasingly (bi-)racialized images of the two major parties magnify the impact of skin
tone on partisanship. That said, it is also consistent with alternative explanations: perhaps the
ANES measure of skin tone contains less random error, due to the fact that it has more variation.
Another sign of the limit of the skin color/partisanship relationship is as follows: even among the
lightest-skinned Latinos in the 2012 ANES, the predicted value for partisanship when control
variables are set to their means is between Independent and Lean Democrat. Light skin tone,
then, might be said to weaken, not reverse, the tendency for Latinos to lean Democratic.
[Table 1 about here]
14
Relationship between Skin Tone and Latino Vote Choice in the 2012 Presidential Election
If light-skinned Latinos are less likely to identify as Democratic, they should be less
likely to vote for Democratic candidates as a result. To examine this possibility, we conduct two
logistic regressions; in both, the dependent variable is vote choice in the 2012 presidential
election (2012 ANES). Only those respondents voting for one of the two major-party candidates
are included. The results (Table 2, Model 1) show that indeed, light skin tone is negatively, and
powerfully, associated with vote choice for Obama in the 2012 presidential election (p<0.01).
The magnitude of the effect is displayed in a predicted probability plot in Figure 3, in which
control variables are set to their means. Here we see that among the darkest-skinned Latinos the
predicted probability of voting for Obama is 0.97, while among the lightest-skinned Latinos the
predicted probability of voting for Obama is 0.43, a decrease of 54 percentage points. Even
restricting the interpretation to those 92% of the Latino respondents at the lighter-skinned end of
the measure yields a large association, from a predicted probability of voting for Obama of 0.87
at the dark end of the scale to a predicted probability of voting for Obama of 0.43 at the light
end, a decrease of 44 percentage points.4
We have argued that light-skinned Latinos are less likely to identify as Democrats and
are therefore less likely to vote for Democratic candidates. If this causal chain is correct, one
observable implication is that the relationship between skin color and vote choice seen here
ought to be attenuated when party identification is included in the model. In order to test this
hypothesis, the analysis of vote choice in the 2012 presidential election in Model 1 of Table 2 is
4 Additional analyses indicate that while associations between light skin tone and vote choice for House and Senate elections are in the expected direction (in which light-skinned Latinos are less likely to vote Democrat), these findings are only marginally statistically significant (p<0.07 in both cases), possibly because the sample size drops substantially due to rolloff.
15
replicated identically except for one thing: party identification is added as an independent
variable (Table 2, Model 2). Indeed, and as expected, the results show that the coefficient on skin
tone is closer to zero in the model that includes party identification, and no longer statistically
significant.
[Table 2 about here]
[Figure 3 about here]
Robustness of Relationships between Latino Skin Tone and Partisanship, Vote Choice
It is important to note that the pattern of results described here extends to a variety of
analytic approaches. First, the findings apply to a range of alternative dependent variables
beyond those presented here; for example, light-skinned Latinos are less likely to feel warmly
toward Obama, as measured in both the pre- and post-election waves of the 2012 ANES. Light-
skinned Latinos are also more likely to think the Obama administration favors blacks, consistent
with the interpretation that racial attitudes play a critical role in Latino partisan preferences
Second, the results are not driven by a relatively small number of respondents at the dark
end of the scale. Up until this point, the darkest-skinned respondents have been included in all
models, although they have sometimes been excluded from estimates of the magnitude of
associations in order to deal with skew in the skin tone variable. But re-estimating the 2012
ANES analyses in Table 1 while excluding the 8% of respondents falling in the darkest five
categories on the eleven-point scale changes little (Appendix 3).
Third, factors related to the survey interview are also important to consider. For example,
due to concerns about potential bias in interviewer measures of respondent skin color (Hannon
16
2014), it is important to note that the findings are almost identical when controlling for
interviewer ethnicity and skin color (n.b., Hill 2002).5 Furthermore, breaking out the 2012 ANES
analyses by respondents interviewing in English and respondents interviewing in Spanish reveals
that both sets of respondents are contributing to the findings, which is important given concerns
about whether English and Spanish survey items are functionally equivalent (Perez 2009).
Additionally, up to this point an alternative hypothesis, that any relationship between skin
color and partisan preferences is due to a confounding relationship between skin color and
country of origin, has been dealt with through including controls for region of heritage. As an
additional approach, we now re-estimate the 2012 ANES analysis presented in Table 1 but
restrict the analysis to those respondents who identify as Mexican, Mexican American, or
Chicano. This supplementary analysis results in limitations on statistical power due to the
decrease in sample size. But the general substantive pattern in which light-skinned Latinos are
less likely to identify as Democrats and to vote for Democrats remains evident. The association
between skin color and party identification, moreover, approaches conventional standards of
statistical significance (p<0.07).
Finally, removing pure Independents, “don’t know” and “refused” responses from the
analysis, on the grounds that the decision of whether to engage with the party system is prior to
the decision of which party to align with (Hajnal and Lee 2011), only strengthens the results
reported here. In other words, the relationship between skin tone and the party identification
scale is strongest among those Latinos who explicitly choose a party.
The Skin Tone Priming Experiment
5 These findings are not presented here, since about 60 respondents had missing values for these variables.
17
We now build on these findings of observational associations between skin tone and
partisan preferences, examining experimental results in order to obtain increased analytical
leverage on the question of causation. Among Latino respondents in the Skin Tone Priming
Experiment, we see no difference in average partisanship across conditions: the mean partisan
score for Latinos who were asked the skin tone question prior to the partisanship question was
indistinguishable from the mean partisan score for those who were asked their skin tone after
being asked their partisanship. However, this null aggregate result masks important
countervailing findings.
Figure 4 displays the marginal effect of assignment to the Priming Condition (y-axis) by
skin tone (x-axis). The marginal effects are only displayed for those Latinos in the six lightest
categories of the skin tone measure, since less than one percent of respondents fell in the four
darkest categories. Among those primed to think about skin tone when answering the
partisanship question, we see a clear pattern of polarization: light-skinned Latinos become more
Republican, as anticipated, and additionally, dark-skinned Latinos become more Democratic.
Both effects are statistically significant (p < 0.05). Furthermore, the slope of the line is
statistically significant; the relationship between skin tone and partisanship grows when Latino
respondents are primed to think about skin tone when answering the partisanship question. This
experimental finding leads to increased confidence that skin tone causes partisan preferences.
[Figure 4 about here]
Assessing the Mechanism: the Role of Racial Attitudes
We have argued that racial attitudes play a key role in the relationship between skin tone
and Latino party identification. Specifically, light-skinned Latinos are more likely than their
18
dark-skinned counterparts to have the option of becoming “honorary whites.” And part and
parcel of affiliating themselves with whites – and distancing themselves from blacks – is
identifying with the party widely perceived to look out for whites’ interests, the Republican
Party, while turning their backs on the party widely perceived to look out for blacks’ interests,
the Democratic Party. If this is correct, than certain empirical consequences follow.
First, we should see a correlation between Latino skin tone and attitudes toward blacks
and whites. That is, lighter-skinned Latinos should be more likely than dark-skinned Latinos to
identify as white and to say positive things about whites, and less likely than dark-skinned
Latinos to identify as black and to say positive things about blacks. As Bonilla-Silva and Dietrich
(2009) put it, “honorary whites” such as light-skinned Latinos should “be in the process of
developing whitelike racial attitudes befitting of their new social position and differentiating
(distancing) themselves from the collective black.” To consider this possibility, we estimate a
series of regression models identical to those in Table 1 except the dependent variables are
measures of racial attitudes rather than party identification and vote choice. The results are
presented in Table 3.
The first column of the table displays the results of a logistic regression analysis in which
the dependent variable is coded “1” if the respondent identifies as White and “0” otherwise.
Consistent with expectations, Latinos with lighter skin are substantially more likely than dark-
skinned Latinos to identify as White. A calculation of the predicted probabilities (in which
control variables are set to their means) reveals that this association is substantial in magnitude;
among those Latinos at the darkest-skinned end of the scale, the predicted probability of
identifying as White is 31 percent, while among Latinos at the lightest-skinned end of the scale,
19
the predicted probability of identifying as White is 78 percent, an increase of 47 percentage
points. This association is statistically significant (p<0.05).
In the second column, meanwhile, we see that light skin tone is negatively associated
with self-identification as Black among Latinos. Here too calculating the predicted probabilities
is necessary, since it is difficult to interpret the magnitude of the effects from a logistic
regression coefficient alone. Among those Latinos at the darkest-skinned end of the scale, the
predicted probability of identifying as Black is 54 percent, while among those Latinos at the
lightest-skinned end of the scale, the predicted probability of identifying as Black approximates 0
percent. In short, and as expected, skin color is strongly associated with how Latinos self-
identify racially.
Indeed, this pattern is evident not only with respect to self-identification but also with
respect to evaluations of Whites and Blacks. The remaining three columns of Table 3 present the
results of models in which the dependent variable is a differential of ratings of Whites and
Blacks (n.b., Hutchings 2009). Respondents rate the work ethic of both Whites and Blacks, for
example, and a measure of the differential between the two is created by subtracting how
hardworking the respondent rates Blacks on a 1 to 7 scale from how hardworking the respondent
rates Whites (n.b., Hutchings 2009). A similar process is used to calculate an intelligence
differential for Whites and Blacks as well. Finally, respondents are asked how warm they feel
toward Whites and Blacks on a 0 to 100 scale; here too a differential is constructed.
As expected, the table shows that light-skinned Latinos are more likely than dark-skinned
Latinos to rate Whites favorably relative to Blacks. Since the regressions have ordinary least
squares specifications and since the variables are mapped onto a 0 to 1 scale, the magnitude of
the association is easy to discern: between 7 and 14 percent of the scale across these three
20
dependent variables. The associations are statistically significant in two of the three cases and
marginally significant in the third, the intelligence stereotype (p<0.07). Furthermore, additional
analyses breaking these White/Black differential measures into these component parts reveal that
evaluations of both Whites and Blacks are contributing to the effects presented here. Consistent
with our argument, light skin is positively associated with Latino identification as White and
positive evaluations of Whites; it is negatively associated with Latino identification as Black and
positive evaluations of Blacks.
[Table 3 about here]
An additional observable implication of the claim that skin tone affects partisanship
through racial attitudes is as follows: the relationship between skin tone and party identification
should be attenuated when racial attitudes are included in the model. To assess this possibility,
we estimate an ordinary least squares regression identical to that in the first column of Table 1
except that the racial attitude measures from Table 3 are also included as independent variables.
The results, presented in Table 4, show that indeed, with racial attitudes held constant, the
relationship between skin tone and partisanship is cut substantially and drops from statistical
significance.
[Table 4 about here]
The Racialized Partisan Images Experiment
We now turn once again to an experimental approach, this time in order to more directly
speak to the causal role of racialized images of the two major parties. Here we compare two
groups of respondents: those in the Racialized News Condition, and those in either the Non-
Racialized News Condition or the Control Condition. This analytic approach is taken to
21
maximize statistical power, since we found no differences between these latter two conditions.
As in the case of the previous experiment, we see no net effect of assignment to experimental
condition on average partisanship. Here too, though, we see important polarizing effects of
racialized images of the two major parties on Latino partisanship.
That is, Figure 5 shows that among Latinos exposed to media messages emphasizing the
changing racial constituencies of the two major parties, light-skinned Latinos become more
Republican, and dark-skinned Latinos become more Democratic. Both of these effects are
statistically significant (p<0.05). Furthermore, the slope of the line is statistically significant; the
relationship between skin tone and Latino partisanship is magnified when Republicans are linked
with Whites and Democrats with Blacks in the news media. This finding increases our
confidence that (bi-)racialized images of the two major parties are a key mechanism underlying
the relationship between skin tone and Latino partisanship.
[Figure 5 about here]
Conclusion
Despite the valuable contributions by previous scholarship, we know much less than we
might about how Latinos situate themselves in the increasingly (bi-)racialized partisan landscape
of contemporary American politics. This project takes some initial steps to address this question.
We argue that Latinos with light skin tone have an opportunity to distance themselves from
Blacks and affiliate with Whites. This decision to affiliate with Whites, we argue, also spurs
affiliation with the party perceived to represent Whites: the Republican Party. Consistent with
this argument, we find an observational relationship between skin tone and partisanship across
three nationally representative datasets – indeed, light-skinned Latinos are not only more likely
22
to self-identify as Republican but also more likely to have voted for the Republican candidate,
Mitt Romney, in the 2012 presidential election. Furthermore, priming skin tone increases its
relationship with partisanship, substantiating the argument that skin tone has a causal effect.
In further support of the argument that light-skinned Latinos are more likely to seek to
affiliate themselves with Whites and the party perceived to represent them, we also present
evidence that light skin tone is associated with unfavorable views of Blacks and favorable views
of Whites. Indeed, an experiment shows that exposure to news media messages linking Whites
with the Republican Party increases the likelihood that light-skinned Latinos identify as
Republican. Latinos face a biracial partisan landscape, and how they navigate that landscape
depends on the interaction of their skin tone and the racialized messages they receive about the
two major parties.
One limitation of the research presented here is that we have not yet addressed an
additional possible route through which skin tone might influence Latino partisanship: dark-
skinned Latinos might be more likely to experience discrimination, leading them to the party
perceived to be more likely to protect Latinos against discrimination, the Democratic Party.
While we have not yet been able to address this possibility thoroughly here, we intend to do so in
a future iteration of this manuscript.
These findings have important implications not only for the study of race and
partisanship but also for the study of race in political science more generally. While the scholarly
consensus is that race is a social construction, in practice many studies treat racial groups, along
with racialized panethnic groups such as Latinos, as fixed, exogenous, and internally racially
homogeneous. Such a decision results not from intellectual sloppiness but from the need for a
meaningful starting point: indeed, it is hard to imagine that scholars could have made so much
23
progress in the study of the relationship between race and partisanship otherwise. But we argue
that in at least one important case, the case of those individuals bounded together in analyses
under the category Latino, there is politically consequential and racially relevant phenotypic
variation within this group that has been missed by previous research. Some Latinos are lighter-
skinned than others, powerfully conditioning their relationship to the partisan structure of the
American political system. By ignoring this critical within-group phenotypic variation among
Latinos, existing scholarship has actually underestimated the extent to which American partisans
are “divided by color.”
24
References
Berinsky, Adam J., Gregory A. Huber, and Gabriel S. Lenz. 2012. "Evaluating online labor
markets for experimental research: Amazon.com's Mechanical Turk." Political
Analysis 20: 351-368.
Bonilla-Silva, Eduardo, and David R. Dietrich. 2009. "The Latin Americanization of US race
relations: A new pigmentocracy." Shades of difference: Why skin color matters (2009):
40-60.
Brooks, Clem, and Jeff Manza. 1997. "Social cleavages and political alignments: US presidential
elections, 1960 to 1992." American Sociological Review: 937-946.
Cain, Bruce E., D. Roderick Kiewiet, and Carole J. Uhlaner. 1991. "The acquisition of
partisanship by Latinos and Asian Americans." American Journal of Political Science:
390-422.
Hutchings, Vincent L. 2009. “Change or More of the Same? Evaluating Racial Attitudes in the
Obama Era.” Public Opinion Quarterly 73: 917-942.
Fox, Cybelle, and Thomas A. Guglielmo. 2012. "Defining America’s Racial Boundaries: Blacks,
Mexicans, and European Immigrants, 1890–19451."American Journal of Sociology 118:
327-379.
Frank, Reanne, Ilana Redstone Akresh, and Bo Lu. 2010. "Latino Immigrants and the US Racial
Order How and Where Do They Fit In?" American Sociological Review 75: 378-401.
Frymer, Paul. 2010. Uneasy alliances: Race and party competition in America. Princeton
University Press.
Green, Donald P., Bradley Palmquist, and Eric Schickler. 2004. Partisan hearts and minds:
Political parties and the social identities of voters. Yale University Press.
25
Hajnal, Zoltan, and Taeku Lee. 2011. Why Americans Don’t Join the Party. Princeton University
Press.
Hannon, Lance. 2014. “Hispanic Respondent Intelligence Level and Skin Tone: Interviewer
Perceptions from the American National Election Study.” Hispanic Journal of
Behavioral Sciences 36: 265-283.
Hill, Mark E. 2002. “Race of the Interviewer and Perception of Skin Color: Evidence from the
Multi-City Study of Urban Inequality.” American Sociological Review 67: 99-108.
Hutchings, Vincent L. 2009. "Change or more of the same? Evaluating racial attitudes in the
Obama era." Public Opinion Quarterly 73: 917-942.
Hutchings, Vincent L., and Nicholas A. Valentino. 2004. "The centrality of race in American
politics." Annu. Rev. Polit. Sci. 7: 383-408.
Iyengar, Shanto, Gaurav Sood and Yphtach Lelkes. 2012. “Affect, Not Ideology: A Social
Identity Perspective on Polarization.” Public Opinion Quarterly 76: 405-431.
Kinder, Donald R., and Allison Dale-Riddle. 2012. End of Race?: Obama, 2008, and Racial
Politics in America. Yale University Press.
Kinder, Donald R., and Lynn Sanders. 1996. Divided by color: Racial politics and democratic
ideals. University of Chicago Press.
Lee, Jongho, and Harry P. Pachon. 2007. "Leading the Way: An Analysis of the Effect of
Religion on the Latino Vote." American Politics Research 35: 252-272.
Lee, Taeku, and Efrén O. Pérez. 2014. "The persistent connection between language-of-
interview and latino political opinion." Political Behavior 36: 401-425.
Luttig, Matthew. 2015. “Obama, Race, and the Republican Landslide in 2010.” Politics,
Groups, and Identities 1: 1-23.
26
MacKuen, Michael B., Robert S. Erikson, and James A. Stimson. 1989. “Macropartisanship.”
American Political Science Review 83: 1125-1142.
Mason, Lilliana. 2015. “I Disrespectfully Agree”: The Differential Effects of Partisan Sorting
on Social and Issue Polarization.” American Journal of Political Science 59: 128-145.
Massey, Douglas, Camille Z. Charles, Garvey Lundy, and Mary Fisher. 2003. "The source of the
river: The origins, aspirations, and values of freshmen at America’s elite colleges and
universities." New York: Russell Sage Foundation.
Pasek, Josh, Tobias H. Stark, Jon A. Krosnick, Trevor Tompson, and B. Keith Payne. 2014.
“Attitudes toward Blacks in the Obama Era: Changing Distributions and Impacts on Job
Approval and Electoral Choice, 2008-2012.” Public Opinion Quarterly 78: 276-302.
Perez, Efren. 2009. “Lost in Translation? Item Validity in Bilingual Political Surveys.” Journal
of Politics 71: 1530-1548.
Piston, Spencer. 2010. “How Explicit Racial Prejudice Hurt Obama in the 2008 Election.”
Political Behavior 32: 431-451.
Slothuus, Rune, and Claes H. de Vreese. 2010. “Political Parties, Motivated Reasoning, and
Issue Framing Effects.” Journal of Politics 72: 630-645.
Tesler, Michael. 2013. “The Return of Old-fashioned Racism to White Americans’ Partisan
Preferences in the Early Obama Era.” Journal of Politics 75: 110-123.
Tesler, Michael. Most Racial: The Growing Racialization of Mass Politics in the Obama Era.
Forthcoming at University of Chicago Press.
Valentino, Nicholas A., and David O. Sears. 2005. "Old times there are not forgotten: Race and
partisan realignment in the contemporary South." American Journal of Political
Science 49: 672-688.
27
Weinberg, Jill D., Jeremy Freese, and David McElhattan. 2014. "Comparing data characteristics
and results of an online factorial survey between a population-based and a crowdsource-
recruited sample." Sociological Science: 292-310.
Williams, David R., Yan Yu, James S. Jackson, and Norman B. Anderson. 1997. “Racial
Differences in Physical and Mental Health Socio-economic Status, Stress and
Discrimination.” Journal of Health Psychology 2: 335-351.
Wimmer, Andreas. 2008. "The making and unmaking of ethnic boundaries: A multilevel process
theory." American Journal of Sociology 113: 970-1022.
Figure 1. Measure of Skin Color in the 2012 ANES
The New Immigrant Survey included a skin color scale developed out of work by Massey and colleagues (2003); this scale was the basis for the measure used in the 2012 ANES. Instructions to interviewers are as follows: “As you know, human beings display a wide variety of physical attributes. One of these is skin color. Unfortunately it is a reality that some respondents may answer questions differently depending on theirs and your skin tone. They may not be comfortable being honest about their opinions, for example, if they are worried they might offend you. In order to detect such discrimination, it is important that the study include a measure of skin color. In order to address the potential bias that skin color may introduce, we will ask you to record the skin color of the Respondent using a Scale of Skin Color Darkness. This is an 11-point scale, ranging from zero to 10, with zero representing albinism, or the total absence of color, and 10 representing the darkest possible skin. The eleven shades of skin color are depicted in a chart, with each point represented by a hand, of identical form, but differing in color. You should be careful to assess the Respondent's skin color regardless of his or her race or ethnicity. It is important that you become familiar with the scale so that you do not access it during the interview. Respondents should never see this scale.” In the analyses presented in this paper, the scale is reversed such that higher values indicate lighter skin color.
Figure 2. Distribution of Skin Tone Among Latinos
Source: 2012 ANES. Only those respondents with non-missing values for control variables in regression analyses (see Table 1) are shown here (N = 431). Including all Latinos (N = 458) does not meaningfully change the distribution. The analysis is unweighted. For the measure of skin tone, see Figure 1.
Figure 3. Effect of Skin Tone on Latino Vote in 2012 Presidential Election Based on Table 1, Model 1
Figure 4. Priming Skin Tone Increases Its Relationship with Latino Party Identification
Source: 2015 MTurk Study. Y-axis represents the effect of assignment to the skin tone priming condition on partisanship (higher values indicate more Democratic partisanship).
Figure 5. Racialized Coverage of the Parties Increases the Relationship between Skin Tone and Latino Partisanship
Source: 2015 MTurk Study. Y-axis represents the effect of assignment to the condition in which the parties are described in racialized terms on partisanship (higher values indicate more Democratic partisanship).
Table 1. Associations Between Latino Skin Tone and Partisanship 1989-90 LNPS 2008 CMPS 2012 ANES Skin Tone -0.09** -0.09* -0.23** (0.04) (0.05) (0.10) Education 0.01 -0.05 -0.10 (0.06) (0.04) (0.07) Age 0.28*** 0.04 0.05 (0.05) (0.04) (0.07) Male -0.03 -0.06** -0.09** (0.02) (0.02) (0.04) Income -0.03 -0.12* 0.07 (0.04) (0.05) (0.08) Income - Missing -0.06 -0.05* -0.04 (0.04) (0.03) (0.06) Own Home 0.04 -0.01 -0.03 (0.02) (0.03) (0.03) Born in U.S. 0.04* -0.01 0.02 (0.02) (0.03) (0.06) Spanish Interview -0.03 0.02* -0.03 (0.03) (0.01) (0.09) Constant 0.55*** 0.85*** 0.93*** (0.09) (0.07) (0.09) N 1766 1390 431 R-squared 0.27 0.10 0.14
*** p<0.001, ** p<0.01, * p<0.05; cell entries are ordinary least squares regression coefficients (standard errors are in parentheses). The dependent variable is the standard seven-point measure of partisanship, in which the highest value indicates self-identification as a Strong Democrat. Higher values of the skin tone variable indicate lighter skin. Models also control for country/region of heritage and religion, and the 2008 CMPS has an additional control for missing values of the age measures, in order to minimize loss of sample size due to listwise deletion. All variables coded 0 to 1. All analyses run using weights for national representativeness.
Table 2. The Relationship between Latino Skin Tone and Vote Choice in the 2012 Presidential Election
(1) (2) Skin Tone -4.95** -3.73
(1.73) (3.18) Party ID (Dem.) 9.48***
(1.65) Education -0.76 0.58
(1.13) (1.31) Age -1.81** -3.09**
(0.59) (1.22) Male -0.22 0.13
(0.51) (0.71) Income 1.06 3.03
(1.32) (1.83) Income – Miss. -0.89 -0.55
(0.72) (0.81) Own Home -0.67 -1.22
(0.53) (1.35) Born in U.S. 1.62* 1.91*
(0.70) (0.83) Spanish Interview 1.68* 4.51***
(0.94) (1.31) Constant 5.23*** -1.58
(1.41) (2.51) N 232 231
R-sq./F-stat. 3.10 6.06 Cell entries are logistic regression coefficients; dependent variable is vote choice in the 2012 presidential election (higher value indicates voting for the Democratic candidate, Obama). See Table 1 for additional notes.
Table 3. Latino Skin Tone is associated with Racial Attitudes
Identify as
White Identify as
Black
Blacks Lazier than
Whites
Blacks Less Intelligent
than Whites
Blacks Rated Less
Warmly than Whites
Skin Tone 2.34* -7.10** 0.14** 0.07 0.14*** (1.01) (2.79) (0.06) (0.04) (0.04)
Education -0.57 0.01 0.00 -0.03 -0.04 (0.70) (0.86) (0.04) (0.02) (0.03)
Age 1.27* 0.22 0.02 0.01 -0.00 (0.54) (1.38) (0.03) (0.02) (0.02)
Male 0.02 -0.18 -0.02 -0.01 -0.01 (0.33) (0.51) (0.02) (0.01) (0.01)
Income 1.74** 1.22 0.06* 0.02 0.01 (0.70) (1.12) (0.03) (0.03) (0.03)
Income – Miss. 1.02* 1.73* -0.01 -0.00 -0.02 (0.44) (0.85) (0.02) (0.02) (0.02)
Own Home -0.20 -1.64** -0.00 -0.01 0.03** (0.28) (0.66) (0.01) (0.01) (0.01)
Born in U.S. 0.71** -0.84 0.00 0.01 0.01 (0.30) (0.57) (0.02) (0.01) (0.02)
Spanish 0.61 0.14 0.05 -0.05* 0.05* Interview (0.50) (0.81) (0.03) (0.02) (0.03) Constant -2.23** 1.77 0.43*** 0.53*** 0.40***
(0.75) (1.92) (0.06) (0.04) (0.04) N 435 435 402 401 396
R-sq./F-stat. 3.07 5.46 0.10 0.07 0.12 Source: 2012 ANES. Column heading indicates dependent variable. Racial identification (first two columns) is a dichotomous variable, and logistic regressions are used; all other models are ordinary least squares regressions. See Table 1 for additional notes.
Table 4. Racial Attitudes Attenuate the Skin Tone/Party ID Relationship. Party ID (Dem.) Party ID (Dem.)
Skin Tone -0.23** -0.14 (0.10) (0.10) Resp. IDs as White 0.04 (0.03) Resp. IDs as Black 0.06 (0.07) Blacks Lazier than Whites -0.02 (0.17) Blacks Less Intelligent than Whites -0.11 (0.15) Blacks Rated Less Warmly than Whites -0.41* (0.17) Education -0.10 -0.13* (0.07) (0.08) Age 0.05 -0.01 (0.07) (0.06) Male -0.09** -0.10** (0.04) (0.04) Income 0.07 0.03 (0.08) (0.09) Income - Missing -0.04 -0.06 (0.06) (0.06) Own Home -0.03 -0.00 (0.03) (0.04) Born in U.S. 0.02 0.04 (0.06) (0.07) Spanish Interview -0.03 0.01 (0.09) (0.10) Constant 0.93*** 1.13*** (0.09) (0.15) N 431 387 R-squared 0.14 0.17
See Table 1 for notes.
Appendix 1. Question Wording, 2012 ANES Skin Color See Figure 1. Ethnicity “Are you Spanish, Hispanic, or Latino?” Hispanic subgroup (heritage) “Are you Mexican, Mexican-American, Chicano, Puerto Rican, Cuban, Cuban-American, or some other Spanish, Hispanic, or Latino group?” Stereotypes
[This question was also asked for whites.]
[This question was also asked for whites.] Warmth “Using the same thermometer scale which you used earlier in the interview, how would you rate:”
[This question was also asked for whites.]
Appendix 2. Racialized Partisanship in the News Experiment
Non-Racialized News Condition
Southerners’ Loyalty to G.O.P. Growing President Obama’s landslide victory in 2008 was supposed to herald the beginning of a new Democratic era. And yet, six years later, there is not even a clear Democratic majority in the country, let alone one poised for 30 years of dominance. It’s not because Mr. Obama’s strong coalition of voters failed to live up to its potential. They again turned out in record numbers in 2012. The Democratic majority has failed to materialize because the Republicans made large, countervailing and unappreciated gains of their own in the South.
Mitt Romney addresses a group of supporters during the 2012 campaign.
From the high plains of West Texas to the Atlantic Coast of Georgia, Southerners opposed Mr. Obama’s re-election in overwhelming numbers. In many counties 90 percent of Southerners chose Mitt Romney.
President Obama speaks in front of a large crowd.
While Southerners have been voting Republican for decades, the large size of the gap was new. Mr. Obama often lost more than 40 percent of Al Gore’s support among voters in the south of the line of the Missouri Compromise. Southern politics are deeply polarized.
Racialized News Condition
Southern Whites’ Loyalty to Republicans Nearing Blacks’ Loyalty to Democrats President Obama’s landslide victory in 2008 was supposed to herald the beginning of a new Democratic era. And yet, six years later, there is not even a clear Democratic majority in the country, let alone one poised for 30 years of dominance. It’s not because Mr. Obama’s strong coalition of black voters failed to live up to its potential. They again turned out in record numbers in 2012. The Democratic majority has failed to materialize because the Republicans made large, countervailing and unappreciated gains of their own among white Southerners.
Mitt Romney addresses a group of mostly white supporters during the 2012 campaign. From the high plains of West Texas to the Atlantic Coast of Georgia, white voters opposed Mr. Obama’s re-election in overwhelming numbers. In many counties 90 percent of white voters chose Mitt Romney; approaching the loyalty of blacks to the Democratic Party; 95% of black voters in the nation supported Mr. Obama.
President Obama speaks in front of a large, predominantly black crowd of supporters.
While white Southerners have been voting Republican for decades, the large size of the gap was new. Mr. Obama often lost more than 40 percent of Al Gore’s support among white voters south of the historically significant line of the Missouri Compromise. Two centuries later, Southern politics are deeply polarized along racial lines. It is no exaggeration to suggest that in these states the Democrats have become the party of African Americans and that the Republicans are the party of whites.
Appendix 3. Same analysis as Table 1 except those with darkest skin tones are excluded
Party ID (Dem.)
Skin Tone -0.22* (0.11)
Education -0.10 (0.07)
Age 0.06 (0.07)
Male -0.09** (0.04)
Income 0.08 (0.08)
Income - Missing -0.04 (0.06)
Own Home -0.03 (0.03)
Born in U.S. 0.01 (0.06)
Spanish Interview -0.04 (0.09)
Constant 0.92*** (0.10)
N 424 R-squared 3.56
Source: 2012 ANES