In November of
2016, Donald Trump shocked the country by defeating Hillary Clinton in the
election for the presidency. The unexpected outcome spurred a debate among
academics over how and why a candidate with no political experience, a long
list of personal scandals, and a dearth of policy knowledge could have won an
electoral contest for the most powerful political office in the country. One
theory is that Trump’s supporters were motivated by what Achen Bartels calls blind
retrospection (Bartels, 2016). Regardless of the quality of the candidate or
the policies of the political parties competing in an election, voters who
believe they are suffering from socioeconomic problems in the months before the
election become more likely to vote for the challenger over a member of the
incumbent party. An alternative theory is that Trump supporters were motivated
by his populist rhetoric during the campaign, which tapped into feelings of
symbolic racism and even old fashioned racism among America’s white
population. Trump frequently spoke
negatively about Blacks, Hispanic immigrants, and Muslims, which may have
earned him votes among whites with strong feelings of racial resentment and
ethnocentric biases. Which of these theories is a better predictor of vote
choice?
To test these theories, I analyzed
data from the 2016 American National Election Survey (ANES), which was conducted
following the presidential election and contained a stratified sample of the
American population. The dependent variable vote choice is binary as votes for
Donald Trump were coded as one and votes for Hillary Clinton were coded as
zero. Votes for the third party
candidates were eliminated from the data, which left a total of 1,171
respondents. Since the dependent
variable vote choice is categorical, this requires the use of a binary logistic
regression model to determine which independent variables had a greater
probability of influencing the outcome of the election.
To measure the independent variable
blind retrospection, I used a question from the 2016 ANES that asked
respondents about the state of the economy. They were given the following five
choices: 1. Very good 2. Good 3. Neither good nor bad 4. Bad 5. Very bad. Self-reported feelings on economic status is
a better measure of blind retrospection than actual unemployment numbers or
some other macro measure of the economy since we are concerned with peoples’
perceptions of reality and not reality itself. It is my hypothesis that
respondents who rate themselves higher on this scale will have a greater
probability of voting for Donald Trump. Furthermore, Achens’ theory of blind
retrospection assumes that voters behave this way due to political ignorance. I
use a question on the ANES survey that asked respondents to rate the extent to
which they followed politics on a scale of one to five with one being always
and five being never. An independent variable measuring the level of education
is also included in the model. The second hypothesis is that lower levels of political
knowledge, when interacted with perceptions of economic status, will make a
voter more likely to vote for Trump due to blind retrospection.
Measuring the impact of the other
independent variable racism was far more challenging as there are several
theories on racism in America and how to measure the concept. The first type of
racism, old fashioned is racism, is the belief that another race is
biologically inferior to one’s own. The political science literature shows that
old fashioned racism was prominent in American history up until the Civil
Rights Era, then old fashioned racism declined after the 1960’s to the point
where the ANES survey stopped measuring it by the 1990’s. However, several
political scientists such as Michael Tessler believe that Barrack Obama’s electoral
victory in 2008 led to the reemergence of old fashioned racism within the
population (Tessler, 2013). Furthermore, many scholars argue that old fashioned
racism didn’t completely disappear as respondents on surveys are aware of the
social repercussions of answering questions on race honestly. Since the ANES
did not directly measure old fashioned racism in 2016, I used the Black Feeling
Thermometer to test whether negative attitudes towards blacks made a voter more
likely to vote for Donald Trump. This is
an imperfect solution. The third hypothesis is that lower scores on the black
feeling thermometer will make voters more likely to vote for Donald Trump.
The second type of racism, symbolic
racism or modern racism, is cultural and not biological in nature. While old
fashioned racism declined in the 1960’s, many political scientists argue that a
new form of racism emerged to replace it as many whites opposed the
redistribution of wealth to black communities, school integration, and
affirmative action policies (Sniderman et al., 1991). The majority of
conservative whites have blamed problems in the black community on a culture of
laziness and entitlement while denying the impact of the legacy of slavery and
Jim Crow on present socioeconomic and political inequalities. Numerous studies
have shown that symbolic racism is correlated with opposition to social welfare
policies (Kinder and Sears, 1981; Desante, 2013) as well as tougher attitudes
towards criminal justice (Weizer, 2004). Symbolic racism has led to the persistence of
black stereotypes among whites, and these attitudes have affected white
political behavior in terms of voting, party identification, and opinion
formation. To measure symbolic racism, I chose five frequently used questions
from the ANES survey (Tarman and Sears, 2005):
1.
‘Generations of
slavery and discrimination have created conditions that make it difficult for
blacks to work their way out of the lower class.’
2.
Are you for or
against preferential hiring and promotion of blacks?
3.
On a scale of one
to seven, how much should the government help blacks?
4.
Do you favor,
oppose, or neither favor nor oppose allowing universities to increase the
number of black students studying at their schools by considering race along
with other factors when choosing students?
5.
‘It’s really a
matter of some people not trying hard enough, if blacks would only try harder
they could be just as well off as whites.’ Agree/disagree.[1]
These questions either have ordinal responses
on a scale of one to five, ordinal responses on a scale of one to seven, or
they only allow the respondent to agree or disagree. With the exception of
question five on work ethic, higher scores on these variables indicate more
symbolically racist attitudes. My fourth hypothesis is that individuals that
score higher on measurements of symbolic racism will be more likely to vote for
Donald Trump.
Finally, ethnic identity should also
influence how voters perceive Donald Trump if his racialized rhetoric
influenced the outcome of the election. Dummy variables were used for blacks,
Hispanics, and Asians while whites were kept as the constant. My fifth hypothesis is whites should be more
likely to vote for Trump than ethnic minorities when controlling for other
variables.
There are also a great variety of
other factors that can influence the outcome of an election, so several control
variables are needed here. Within the political science literature, party
identification is seen as one of the strongest variables that influence voter
choice (Druckman, Petersen, and Slothus, 2013). Dummy variables are used for
Republican and Democrat, and the constant is used for independents. Political ideology also influences voter
choice. A question is used from the 2016 ANES survey that asks respondents to
rate whether they are liberal or conservative on a scale of one to seven with
seven being the most conservative. A
standard set of demographic variables are also used in the logit model,
including age, gender, income, education, the importance of religion, and
marital status. A question that measures attitudes towards authoritarianism is
also included.
Two
logistical regression models were conducted, one of which produces coefficients
and another which produces the odds ratio for each variable. The results are as
follows:
|
Trump Vote
|
|
|
|
|
|
Coefficients
|
Robust Standard Errors
|
|
|
|
|
|
|
|
Republican
|
1.58
|
(0.35)
|
***
|
|
Democrat
|
-2.23
|
(0.35)
|
***
|
|
Ideology
|
0.56
|
(0.13)
|
***
|
|
Female
|
-0.19
|
(0.25)
|
|
|
Age
|
0.00
|
(0.01)
|
|
|
Married
|
0.53
|
(0.24)
|
*
|
|
Education
|
-0.42
|
(0.27)
|
|
|
Income
|
-0.02
|
(0.02)
|
|
|
Religious Attitudes
|
0.50
|
(0.26)
|
|
|
Authoritarianism
|
0.14
|
(0.09)
|
|
|
Political Knowledge
|
-0.02
|
(0.12)
|
|
|
Black
|
-1.85
|
(0.66)
|
**
|
|
Hispanic
|
-1.75
|
(0.45)
|
***
|
|
Asian
|
-1.37
|
(0.66)
|
*
|
|
Other Races
|
-0.78
|
(0.38)
|
*
|
|
Economic Status
|
0.84
|
(0.13)
|
***
|
|
Black Thermometer
|
0.00
|
(0.01)
|
|
|
Uni Affirmative Action
|
0.16
|
(0.08)
|
|
|
Black Welfare
|
0.14
|
(0.09)
|
|
|
Affirmative Action
|
0.27
|
(0.38)
|
|
|
Attitudes on Slavery
|
0.26
|
(0.10)
|
*
|
|
Black Work Ethic
|
-0.24
|
(0.12)
|
*
|
|
Constant
|
-6.52
|
(1.41)
|
|
|
Pseudo R-Squared
|
0.77
|
|
|
|
Observations 1,171
*p<.05
|
|
|
|
|
**P<.01
|
|
|
|
|
***p<.001
|
|
|
|
|
|
|
|
|
|
Trump Vote
|
|
|
|
|
|
Odds Ratio
|
Robust Standard Errors
|
|
|
|
|
|
|
|
Republican
|
4.84
|
(1.72)
|
***
|
|
Democrat
|
0.11
|
(0.04)
|
***
|
|
Ideology
|
1.75
|
(0.23)
|
***
|
|
Female
|
0.83
|
(0.21)
|
|
|
Age
|
1.00
|
(0.01)
|
|
|
Married
|
1.70
|
(0.42)
|
*
|
|
Education
|
0.66
|
(0.18)
|
|
|
Income
|
0.98
|
(0.02)
|
|
|
Religious Attitudes
|
1.65
|
(0.43)
|
|
|
Authoritarianism
|
1.15
|
(0.11)
|
|
|
Political Knowledge
|
0.98
|
(0.12)
|
|
|
Black
|
0.16
|
(0.10)
|
**
|
|
Hispanic
|
0.17
|
(0.08)
|
***
|
|
Asian
|
0.25
|
(0.17)
|
*
|
|
Other Races
|
0.46
|
(0.18)
|
*
|
|
Economic Status
|
2.31
|
(0.31)
|
***
|
|
Black Thermometer
|
1.00
|
(0.01)
|
|
|
Uni Affirmative Action
|
1.17
|
(0.09)
|
|
|
Black Welfare
|
1.15
|
(0.10)
|
|
|
Affirmative Action
|
1.31
|
(0.50)
|
|
|
Attitudes on Slavery
|
1.29
|
(0.13)
|
*
|
|
Black Work Ethic
|
0.79
|
(0.09)
|
*
|
|
Constant
|
0.00
|
(0.00)
|
|
|
Pseudo R-Squared
|
0.77
|
|
|
|
Observations 1,171
*p<.05
|
|
|
|
|
**P<.01
|
|
|
|
|
***p<.001
|
|
|
|
The McFadden pseudo-r squared result
of .77 shows that the independent variables explain 77 percent of the variation
in the respondent vote choice. As expected, the strongest independent variable
is party identification as Republicans were 4.84 times more likely than
Independents to vote for Trump. Furthermore, a one point increase on the
ideological scale towards conservativism made the respondent 1.7 times more
likely to vote for Trump.
The results also show relatively
strong support for part of the blind-retrospection hypothesis. Self-reported
economic status is one of the strongest independent variables. A one unit
increase in the ordinal scale measure towards economic dissatisfaction made the
respondent 2.31 times more likely to vote for Trump.
However,
the results for political knowledge did not produce statistical significance. Furthermore,
a separate regression analysis was conducted interacting economic status with
political knowledge but produced results in the opposite direction as expected.
While Trump voters were voting retrospectively, they were not doing so blindly.
They were regularly following the news and not basing their opinion purely on
their personal status.
Furthermore, the results show much
weaker support for the racism hypothesis. While ethnic minorities were far more
likely to vote against Donald Trump with a high degree of statistical
significance, most of the measurements of old fashioned and symbolic racism did
not produce statistically significant results. All of the coefficients are in
the right direction, but only the questions on attitudes towards slavery and
black work ethic are statistically significant at a P-level below .05. While
voters were polarized based on racial identity, the effects of symbolic racism
and old fashioned racism were more limited.
There are several limitations to
this study. For one, the measurements of racism do not measure subconscious
racism nor did they account for the fact that survey respondents may have not
been honest in their responses to the questions on race. Furthermore, it is
possible that tensions between blacks and whites were not as salient for this
election like in 2008. Measurements of ethnocentrism as well as attitudes
towards Muslims and Hispanic immigrants may produce more statistically
significant results. Lastly, this study does not determine why opinions on
personal economic status had such a strong effect on vote choice. Do these
opinions match real deteriorations in the status of living for many Trump
voters, or are these opinions shaped by the conservative media they are reading
and watching? The lack of significance for the income variable in the
regression analysis indicates that the later may be truer than the former.
Work Cited
Bartels, Achen. 2016. Democracy
for Realists. Princeton University
Press: Princeton.
DeSante, Christopher D. 2013. “Working Twice as Hard to Get
Half as Far: Race, Work Ethic, and America's Deserving Poor.” American Journal of Political Science,
57(2): 342-356.
Druckman, James,
Erik Peterson, and Rune Slothus.
2013. “How Elite Partisan
Polarization Affects Public Opinion Formation,” American Political Science Review, 107: 57-79.
Kinder, Donald and David Sears. 1981. “Prejudice and
Politics: Symbolic Racism versus Racial Threats to the Good Life," Journal of Personality and Social Psychology,
40(3): 414-431.
Nail, Paul R., Helen C. Harton, and Brian P. Decker. 2003.
"Political Orientation and Modern versus Aversive Racism: Tests of Dovidio
and Gaertner's (1998) Integrated Model." Journal of Personality and Social Psychology, 84(4): 754.
Scaglion, Richard and Richard Condon. 1980. “The Structure of
Black and White Attitudes towards Police,” Human
Organization, 39(3): 280-283.
Schlesinger,
Joseph. 1985. “The New American
Political Party,” American Political
Science Review, 79: 1152-1169.
Sniderman, Paul M., Thomas Piazza, Philip E. Tetlock, and Ann
Kendrick. 1991. “The New Racism.” American
Journal of Political Science, 35(2): 423-447.
Taber, Charles, and
Milton Lodge. 2006. “Motivated Skepticism in the Evaluation of
Political Beliefs,” American Journal of
Political Science, 50: 755-769.
Tarman, Christopher and David Sears. 2005. “The
Conceptualization and Measurement of Symbolic Racism.” The 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. 2013. “The Return of Old-Fashioned Racism to
White Americans’ Partisan Preferences in the Early Obama Era,” The Journal of Politics, 75(1): 110-123.
“User’s Guide and Codebook for the ANES 2016 Time Series
Study,” http://www.electionstudies.org/studypages/anes_timeseries_2016/anes_timeseries_2016_userguidecodebook.pdf, 2016.
Weizer, Ronald. 2004. “Race and Perceptions of Police
Misconduct,” Social Problems, 51(3):
305-25.
Zaller, John. 1992. The Nature and Origins of Mass Opinion.
Cambridge University Press: Cambridge.
[1] “User’s Guide and Codebook for the ANES 2016 Time
Series Study,” http://www.electionstudies.org/studypages/anes_timeseries_2016/anes_timeseries_2016_userguidecodebook.pdf, 2016.
