American Empire - Coursework 2 - Voting Behaviours Report

You might also like

Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 11

A report on the relationship between one’s education level and the

presidential candidate one voted for in the aftermath of the 2016


US presidential election.
Hypothesis
Operational: People who have higher levels of educational achievement are less likely to
have voted for Donald Trump in the 2016 US presidential election.
Null: People who have higher levels of educational achievement are not less likely to have
voted for Donald Trump in the 2016 US presidential election

Introduction
The dataset used in the analysis for this report was produced by the Pew Research Centre
and consists of public thoughts and opinions related to the 2016 US Presidential election, as
well as issues relating to the election and census information. The data was collected in
December 2017, a year after the election and so gives interesting insight to opinions soon
after the election. Questions posed in the survey include how respondents voted in the
2016 election, if they would have voted for the same candidate a year on, the primary news
source for respondents and how each party prioritises the interests of certain social groups,
amongst other issues. Census data collected in the survey includes age, gender, educational
attainment, health insurance status and social media usage, amongst other things. To
analyse this data, IBM SPSS Statistics was used. This software allows the user to perform
various tasks to establish a connection between two different variables., as well as how
significant this connection is, if at all. When used with tests such as the chi Squared tests
and Cramer’s V test, SPSS lets the user create graphs to compare and display data. To
analyse this data, I chose two variables from the dataset mentioned above, one
independent and one dependent. The independent variable I chose was education level,
with six answer options from “less than high school” to “postgraduate” as well as a “don’t
know/refused” option. The dependent variable I chose (as labelled in the dataset) was “In
the 2016 presidential election who did you vote for?” with answer options allowing
respondents to choose one of the four main candidates; Donald Trump for the republican
Party, Hillary Clinton for the Democratic Party, Gary Johnson for the Libertarian Party, Jill
Stein for the Green Party; as well as an “other” option and a “not sure” option. The purpose
of my research in this report is to either accept or reject my previously mentioned
hypotheses using IBM SPSS to statistically analyse the PEW data, while also drawing on my
own knowledge and other literature on the subject. There will be sections on
contextualisation of my variables, presentation of data, data analysis, conclusions and
recommendations for future research.

Contextualisation of Variables
The purpose of this report is to examine the link between survey respondents’ educational
attainment and who they chose to vote for in the US presidential election, specifically
Donald Trump, the successful candidate. The independent variable I decided to analyse is
educational attainment, with particular focus on the “associate degree” answer option as
this shows respondents have completed an undergraduate degree, however it is important
to note the “postgraduate” option as well, as these respondents are among some of the
most educated in modern society. The dependent variable I have chosen to analyse was the
question asking which presidential candidate the respondent voted for in the 2016 election,
with particular focus on those who voted for Donald Trump, as he was a highly controversial
candidate but went on to win the election. This was a surprise for many people with Trump
being such an unlikely character, especially since the exit polls in the run up to the election
were forecasting a Clinton win. The New York Times reported in 2016 that only one exit poll
out of ten had predicted a Trump victory, even only with a +2-point margin, the lowest of all
polls (New York Times, 2016). This highlighted a massive divide between the two major
political parties in the US, in both political ideology and voting trends which only seems to
be becoming more polarized. Usually, people who feel strong affiliation to the Democratic
party tend to treat everyone equally regardless of their political affiliation, while people
feeling strong affiliation to the Republican party tend to show significant in-group
favouritism. However, since the election, this has changed, with the democrats feeling a
negative shock and therefore causing them to feel less generosity towards Republicans,
while Republicans, no longer feeling the need to practice in group favouritism to be more
generous to everyone, changing the political climate in the USA today (Oc et al. 2018). The
link between education levels and turnout has been explored before, with more educated
citizens generally being considered more engaged with society and therefore are more likely
to vote (Rosenstone & Wolfinger, 1980), with 26% of college educated voters being
consistently liberal when they vote and a further 21% being mostly liberal (Suls, 2016) – so if
college educated voters are not only more likely to vote, but are more likely to vote
liberally, how did incendiary Republican Donald Trump win the election? This raises more
questions about the changing political opinions of US society, allowing for a more nuanced
analysis of why typical voting behaviour may be changing and why more traditional
explanations for voting behaviours may no longer serve a purpose. This report will therefore
statistically analyse education achievement levels in voters of the 2016 presidential election
and the hypotheses will be tested in the context of changing US political climates and
opinions whilst also allowing for questioning of more traditional explanations of voting
behaviour.

Presentation of Data
The data displayed in this section was collected by Pew Research Center, with the two
variables being taken directly from the 2017 survey on the 2016 US Presidential election.
The data has been collated into bar charts and graphs using IBM SPSS.

Cumulative
Frequency. Percent
Percent
Valid
Percent
Valid Less than 130 3.1 3.1 3.1
high school
High school 506 12.1 12.1 15.2
graduate
Some college 923 22.1 22.1 37.3
Associate 414 9.9 9.9 47.2
degree
College 1157 27.7 27.7 74.8
graduate
Postgraduate 25.2 25.2
1053 100.0
Total 100.0 100.0
4183

Figure 1: Frequency table produced using IBM SPSS Statistics, displaying the data
measuring the educational attainment of survey respondents. (Pew Research Center,
2017).

Figure 2: Bar chart produced using IBM SPSS Statistics, displaying the data measuring the
educational attainment of respondents. (Pew Research Center, 2017).

The independent variable was recoded into a new variable with less values. The variable has
now been recoded into three possible values, with “less than high school” and “high school
diploma” being merged into one value: “high school or less”. The values “associate degree”
and “some college” have also been merged into one value: “some college”. The values
“college graduate” and “postgraduate” have been merged into the final value “college
graduate+”. The data has been recoded to allow for stronger analysis as the new values
allow for better distinction between educational levels. This also makes the results more
suitable for the Chi Squared and Cramer’s V tests as there is a higher count in each value
category.

Frequency Percent Valid Cumulative


Percent Percent
Valid High School 636 15.2 15.2 15.2
or Less
Some 1337 32.0 32.0 47.2
College
College 2210 52.8 52.8 100.0
Graduate+
Total 4183 100.0 100.0

Figure 3: Frequency table produced using IBM SPSS Statistics, displaying the data
measuring the educational attainment of survey respondents. (Pew Research Center,
2017) The data has been recoded and placed into three values.

Figure 4: Bar chart produced using IBM SPSS Statistics, displaying the data measuring the
educational attainment of respondents. (Pew Research Center, 2017). The data has been
recoded and placed into four values.

Frequency Percent Valid Cumulative


Percent Percent
Valid Donald 1479 35.4 40.0 40.0
Trump, the
Republican
Hillary 1850 44.2 50.0 90.0
Clinton, the
Democrat
Gary 161 3.8 4.4 94.3
Johnson, the
Libertarian
Party
candidate
Jill Stein, the 53 1.3 1.4 95.8
Green Party
candidate
Voted for 126 3.0 3.4 99.2
none/Other
Refused 30 .7 .8 100.0
Total 3699 88.4 100.0
Missing System 484 11.6
Total 4183 100.0
Figure 5: Frequency chart produced using IBM SPSS Statistics, displaying the data
recording which presidential candidate respondents voted for in the 2016 presidential
election (Pew Research Center, 2017).

Figure 6: Bar chart produced using IBM SPSS Statistics, displaying the data recording
which presidential candidate respondents voted for in the 2016 presidential election (Pew
Research Center, 2017).

The dependent variable has been recoded to allow focus on just the two major candidates
in the election, Trump and Clinton. This makes the results more concise and better
highlights the distinction between the two. This also makes the results more suitable for the
Chi Squared and Cramer’s V tests as there is a higher count in each value category.

Frequency Percent Valid Cumulative


Percent Percent
Valid Donald 1479 35.4 40.3 40.3
Trump
Hillary 1850 44.2 50.4 90.7
Clinton
Other 340 8.1 9.3 100.0
Total 3669 87.7 100.0
Missing System 514 12.3
Total 4183 100.0

Figure 7: Frequency chart produced using IBM SPSS Statistics, displaying the data
recording which presidential candidate respondents voted for in the 2016 presidential
election (Pew Research Center, 2017). The data has been recoded and placed into three
values.

Figure 8: Bar chart produced using IBM SPSS Statistics, displaying the data recording
which presidential candidate respondents voted for in the 2016 presidential election (Pew
Research Center, 2017). The data has been recoded and placed into four values.

Recoded Total
Educational
Attainment
High School Some College
or Less College Graduate
or Higher
Presidential Donald Expected 188.7 457.5 832.8 1479.0
Election Trump Count
Vote
% within 53.4% 48.9% 32.6% 40.3%
Recoded
Educational
Attainment
Hillary Expected 236.0 572.3 1041.7 1850.0
Clinton Count
% within 40.0% 41.9% 57.5% 50.4%
Recoded
Educational
Attainment
Other Expected 43.4 105.2 191.5 340.0
Count
% within 6.6% 9.2% 9.9% 9.3%
Recoded
Educational
Attainment
Total Expected 468.0 1135.0 2066.0 3669.0
Count
% within 100.0% 100.0% 100.0% 100.0%
Recoded
Educational
Attainment

Figure 9: Crosstabulation table displaying data collected by Pew Research Center


measuring education levels and how people voted in the 2016 presidential election. This
table shows which percentage of education level voted for each presidential candidate.
(Pew Research Center, 2017).

Chi-Square Tests
Value Df Asymptotic
Significance (2-
sided)
Pearson Chi-Square 122.127a 4 .000
Likelihood Ratio 122.329 4 .000
Linear-by-Linear 82.467 1 .000
Association
N of Valid Cases 3669
0 cells (0.0%) have
expected count less
than 5. The
minimum expected
count is 43.37.

Figure 10: Chi -Square statistical test for independence produced using IBM SPSS Statistics.
This test is a comparison of the actual data and the data that would be expected if there
was no relationship between the two variables. A relationship is said to be significant if it
has an asymptotic significance of .05 or smaller. (Pew Research Center, 2017).

Value Approximate
Significance
Nominal by Nominal Phi .182 .000
Cramer's V .129 .000
N of Valid Cases 3669

Figure 11: Cramer’s V statistical test produced using IBM SPSS Statistics. This test is used
to measure the strength of the relationship between two variables. The relationship is
measured on a 0-1 scale with 0 meaning no correlation whatsoever and 1 meaning the
relationship is perfectly correlational. (Pew Research Center, 2017).

Analysis
When analysing the dataset, the Chi square test of independence (figure 10) showed a score
of 0.000. This means that there is a relationship between educational attainment and how
one voted in the 2016 US presidential election, as any score below 0.05 shows correlation –
this means we can reject the null hypothesis. However, when we use the Cramer’s V test
(figure 11) which tests the strength of said relationship, we see that the correlation is not
particularly strong with a score of just 0.129 on the test. This puts the result in the “weak
association” category as any score between 0-0.2 is said to represent a weak association
between the two variables. We can still accept our operational hypothesis as there is some
association, however the weak score means we need to investigate further as other factors
are clearly contributing to which presidential candidate one voted for in the 2016 elections.

The relationship shows that generally, better educated people are more likely to have voted
for the Democratic Party candidate, Hillary Clinton (figure 9). This is because the well-
educated tend to be socially liberal and have “down the line liberal views on social,
economic and environmental matters” (Gross, 2016) and with the Democrats today being
the party of civil rights and social progression, this explains why those with more education
would vote for the Democratic candidate. This is reinforced by the fact that those citizens
with higher social and economic status participate more in politics – with one possible
measure of this social capital being education levels (Wolfinger & Rosenstone, 1980).

However, when we take a specific look at other factors contributing to why one may have
voted for each particular presidential candidate, we see the difference in education levels
becomes even more complicated. For example, when examining educational differences via
race, we see that among non-white people in the USA it held little bearing on who they
voted for in the election (Pew research centre, 2018). While overwhelmingly non-white
people generally voted for Clinton, the difference between the voting percentage of college
graduates and non-college graduates was less than 10 points, for both Clinton and Trump.
This shows that, perhaps for these people, their race and the candidate’s stance on racial
issues was a more deciding factor when choosing who got their vote. This is reinforced by
the fact that historically, the Democratic Party has been the party of African-American
people in the USA. This stems from the early 20th century when the Democrats – aided with
Roosevelt’s New Deal – could be seen as “the party that pushed for social services” (Harris,
2018) and so gained the support of African-Americans at the time who were suffering due
to Jim Crow laws and were often poorer than their white counterparts as a result of this.
Reinforced by the civil rights movement in the 1960s and most recently the election of
Barack Obama as the first African-American president, the Democratic Party came to be
known as “the party of progressive, racial politics” (Harris, 2018). On the contrary to this,
when we take a similar look at the same figures for white people in the USA, we see the
‘diploma gap’ is noticeable and significant. Among white Americans who voted for Clinton
there was a 27-point difference between those who were college graduates and those who
were not. This is almost equal for those who voted for Trump, as there was a 26-point
difference between those who were college graduates and those who were not (Pew
research centre, 2018). This shows that race was indeed an important factor for those
voting in the 2016 presidential election as it shows that for non-white people education
level is not the most pertinent reason for how they voted, however, it plays a much more
important role in explaining why white people voted how they did. In fact, it was white
women in particular who got Trump into the white house with 53% of all white women
voting Republican (Jaffe, 2017).

This election was different to the ones that came before it as it featured an extremely
polarised celebrity candidate in the form of Donald Trump. This is sure to have had an effect
on the voting behaviours we usually see in the USA and it is in fact Trump’s celebrity status
and his lack of critical political discourse that could explain this effect. We have seen that
generally non-white Americans followed the pattern seen in many elections before 2016 in
term of how they voted – in fact, black women had the highest voting percentage in the
whole country with 94% of all black women voting for Hillary Clinton (O’Neal, 2016).
Instead, it is white voters – particularly those without a college education – who seemed to
have behaved differently when voting in this election. In the past, working class, rural
America have voted for the Democratic party as it tends to be the party of the poor and the
working class. However, Donald Trump and his rejection of the “political elite” seemed to
tap into the class resentment felt by many rural white voters and use this to his advantage
by seemingly giving an outlet for their anger and a solution to their problems. It is here that
we see the relationship between education levels and who one voted for in the presidential
election as being significant, just not in the way we may have first imagined – as it was those
with lower education levels that strayed from tradition, not the more well educated.
However, it must also be noted how intersectional these factors in voting behaviour are,
with race, gender and education levels often inextricably linked and so while there are
relationships between all of these factors and why people vote a certain way, often one
factor alone cannot decide how a person will vote.

Conclusion
On balance, it is obvious that there is certainly a relationship between a person’s education
level and which presidential candidate they voted for in the 2016 elections, however it is by
no means a flawless example. The hypothesis builds on the common theory that those with
higher education levels will have greater political participation and prompts an investigation
into when these citizens do vote, how they do so. Traditionally, the common explanation for
this relationship has been that generally better educated people tend to be more politically
aware and also more liberal, therefore those with higher education levels would be more
likely to vote for the Democratic candidate. While that is true, this election in particular was
unique as it showed a shift in voting behaviour in which many poorer, less educated
Americans voted for Donald Trump, due to the unique celebrity, “man of the people”, anti-
Washington-elite nature of his campaign, despite usually being a group in society that was
less likely to vote. However, while the relationship between these two variables cannot be
denied, it must also be said that the relationship is dependent upon and affected by outside
factors such as race, gender and the presence of a polarising candidate amongst others. This
exemplifies how one factor can virtually never completely control a citizen’s vote and
instead the decision is usually made up of a myriad of reasons.

Recommendations for Future Research


 Further study into the diploma gap and why this disproportionately affects white
Americans when compared to non-white Americans, and if differences can also be
seen between more advanced higher education such as masters and PhDs.
 The dataset was comprised of less than 4000 people so it would be much improved if
a bigger sample size was used in further research, to get a more accurate view of the
opinion of the general public.
 Another analysis of more than two variables to possibly try and measure the link
between the various factors that go into a citizen’s vote.
 The same study but on the other runners for the presidential nomination for each
party i.e. which Republican candidate republican voters identified with most and
why. This would highlight what is appealing about certain politicians and by
extension certain policy, revealing more about the electoral.

Bibliography
Gross, N. (2016). Why are the Highly Educated so Liberal? The New York Times. [online]
Available at: https://www.nytimes.com/2016/05/15/opinion/why-are-the-highly-educated-
so-liberal.html [Accessed 1 May 2019].

Harris, A. (2018). America Is Divided by Education. The Atlantic. [online] Available at:
https://www.theatlantic.com/education/archive/2018/11/education-gap-explains-
american-politics/575113/ [Accessed 4 May 2019].

Jaffe, S. (2017). Why Did a Majority of White Women Vote for Trump?. New Labor Forum,
[online] 27(1). Available at:
https://journals.sagepub.com/doi/full/10.1177/1095796017744550 [Accessed 3 May 2019].

New York Times (2016). Latest Election Polls 2016. The New York Times. [online] Available
at: https://www.nytimes.com/interactive/2016/us/elections/polls.html [Accessed 3 Apr.
2019].

Oc, B., Moore, C. and Bashshur, M. (2018). When the tables are turned: The effects of the
2016 U.S. Presidential election on in-group favouritism and out-group hostility. PLOS ONE,
[online] 13(5). Available at:
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0197848 [Accessed 8
Apr. 2019].

O'Neal, L. (2016). The 53 Percent Issue. The Undefeated. [online] Available at:
https://theundefeated.com/features/black-women-say-white-feminists-have-a-trump-
problem/ [Accessed 4 May 2019].

Pew Research Centre (2018). An Examination of the 2016 Electorate Based on Validated
Voters. [online] Pew Research Center for the People and the Press. Available at:
https://www.people-press.org/2018/08/09/an-examination-of-the-2016-electorate-based-
on-validated-voters/2-11-2/ [Accessed 3 May 2019].

Suls, R. (2016). Educational divide in vote preferences on track to be wider than in recent
elections. [Blog] Fact Tank. Available at: https://www.pewresearch.org/fact-
tank/2016/09/15/educational-divide-in-vote-preferences-on-track-to-be-wider-than-in-
recent-elections/ [Accessed 11 Apr. 2019].

Wolfinger, R. and Rosenstone, S. (1980). Who votes? New Haven: Yale Univ. Pr.

https://www.people-press.org/2018/08/09/an-examination-of-the-2016-electorate-based-
on-validated-voters/2-12-2/

https://www.nytimes.com/interactive/2016/us/elections/polls.html

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0197848

You might also like