Introduction Through this project, we examined immigration discourse as it related to the 2016
presidential debates. Our goal was to determine how candidates from both sides of the
political spectrum coded their language when discussing concepts like immigration
reform, amnesty, and legal repercussions. We did this by mining the corpus through
manual textual analysis and supplemented our own findings by using collocation and topic
modeling tools to identify commonly used words or phrases, and then analyzed these
results in terms of our larger data set. Our goal for this project was not to make
judgements or draw conclusions based on implicit material; rather, we sought to sort
through and present textual data to be used in further research projects. Within
this discourse, we divided up types of immigration based on status, i.e. we specified
"illegal" and "legal" immigration. We also associated utterances with certain attitudes,
specifically referring to candidates' stances on immigration in pro, anti, and neutral
terms. For longer excerpts, we analyzed tropes, identifiying if certain utterances
referred to immigration in terms of "values", "security", or "economics". Within those
tropes, we specified subtypes, which we unfortunatly did not have sufficient time to
consider in our data analysis.
Collocation
Bush
Word
Count
people
20
can
14
country
13
need
13
fix
11
million
11
world
11
applause
10
get
10
strategy
9
will
9
legal
8
percent
8
president
8
problem
8
Carson
Word
Count
people
24
us
15
know
11
need
9
policies
8
coming
7
going
7
one
7
want
7
way
7
applause
6
country
6
now
6
probably
6
can
5
CHRISTIE
Word
Count
president
6
attorney
5
country
5
general
4
us
4
enforce
3
law
3
officers
3
police
3
states
3
time
3
allows
2
back
2
cities
2
done
2
Clinton
Word
Count
think
29
reform
24
will
20
comprehensive
19
get
19
people
17
president
17
want
17
border
15
know
15
said
15
country
12
work
12
good
11
undocumented
11
Cruz
Word
Count
will
27
amnesty
23
going
20
law
19
bill
18
people
18
can
17
illegal
17
know
16
border
15
applause
14
donald
14
problem
14
supported
13
many
12
Fiorina
Word
Count
problem
8
25
4
arduous
4
ask
4
border
4
democrats
4
go
4
just
4
meanwhile
4
nothing
4
remains
4
say
4
solve
4
solved
4
will
4
Kaine
Word
Count
donald
8
trump
8
will
6
want
5
hillary
4
million
4
people
4
add
3
apologize
3
based
3
enforcement
3
judge
3
nation
3
said
3
16
2
Kasich
Word
Count
people
4
america
3
want
3
believe
2
can
2
come
2
country
2
doors
2
fact
2
go
2
going
2
just
2
lock
2
look
2
problem
2
O'Malley
Word
Count
people
15
applause
10
reform
9
american
8
country
8
need
8
wages
8
can
7
go
7
new
7
america
6
comprehensive
6
economy
6
going
6
us
6
Paul
Word
Count
marco
11
border
8
defend
8
security
8
think
8
coming
7
country
5
forward
5
opposed
5
us
5
bill
4
deal
4
islam
4
made
4
mistake
4
Pence
Word
Count
donald
11
trump
10
country
9
said
9
people
8
border
6
criminal
6
just
6
security
6
want
6
aliens
5
american
5
going
5
reform
5
senator
5
Rubio
Word
Count
people
79
going
49
issue
39
will
39
system
33
now
29
american
27
country
24
way
23
years
23
illegal
22
first
21
support
21
control
20
see
19
Sanders
Word
Count
people
15
applause
13
country
13
bill
8
president
8
reform
8
think
8
united
8
know
7
need
6
states
6
will
6
families
5
going
5
hard
5
Trump
Word
Count
country
20
illegal
17
people
16
said
16
many
15
will
14
think
13
come
11
going
11
say
11
wall
11
just
10
back
9
coming
9
applause
8
Walker
Word
Count
border
4
actually
3
people
3
america
2
american
2
listened
2
need
2
said
2
secure
2
system
2
abbott
1
acknowledged
1
across
1
amnesty
1
applause
1
Webb
Word
Count
country
2
family
2
thousands
2
2007
1
actually
1
amendment
1
american
1
bill
1
boat
1
borders
1
camps
1
communists
1
cornell
1
defining
1
dream
1
Although the collocation analysis generated laundry lists of words that were used
habitually by the candidates, those lists were abbrebviated by the researchers for
a number of reasons. Namely, the number of repetitions associated with words
quickly receded as the list grew progressively longer, and a handful of candidates
were not in enough debates to permit a substantial data return in regard to their
most frequently used terms.
From the lists of words that remain, interesting points still arise in terms of
word repetitions and the implications behind the candidates' most frequently used
words. For instance, Donald Trump and Hillary Clinton, the two candidates in the
general election (and thus two of the candidates with the largest pool of data in
our corpus), have distinctly different words at the tops of their lists that
mirror surprisingly well their policies and views. Clinton's three most commonly
used terms were "think", "reform", and "will," which
align well with the overarching themes seen in the Democratic debates and
policies. As with many candidates, the need for comphrehensive immigration
"reform" was discussed at length by Clinton, with two other future-leaning words
assuming the first and third position on her list. With Trump, a candidate who was
known for his rhetoric on soveriegnty and strength, used the terms
"country", "illegal", and "people" most frequently
in his utterances. His most ritually employed term, country, reflects his chief
argument on rectifying the resounding issues of illegal immigration -- the need to
fortify the American borders and policy against potential erosion from influxes of
undocumented immigrants.
MALLET
We began the topic modeling portion of this project by identifying forty topics with
the text-mining tool MALLET. We only took sections in which immigration was discussed
into consideration, so as to optimize the topic modelling technology to our central
research questions. Thus, we naturally did not consider sections of the debate that
were not included in our study. We ultimately launched the program with forty topics
because we found that this number provided clear sub-topics within immigration that
were relatively easy to identify. We made our own, non-computationally assisted
judgements on the topics, so these informed judgement calls are imperfect. Some
topics were difficult to decipher or irrelevant to our project (potentially due to
the fact that, while we considered any utterance or phrase grouping that included
some keyword, those words were used in a host of contexts over the span of all the
debates), so we did not include them.
These are the relevant sub-topics, divided into larger topics:
Policy: justice, issues, social welfare, laws
International: terrorism, militay, foreign policy,
geopolitics America First: idealism, middle-class, nationalism
Crime: drugs, gun control Economy: economic growth, working america, economy/taxes Other
topics included in the discourse related to references to specific candidates,
references directly to the presidential race or the debate, and other tangentially
related topics including "corporations", "flint crisis",
"police brutality", "obamacare", and "women's rights".
MALLET was able to sort out these topics from the immigration discourse, but they do
not seem directly relevant to this larger project.
Policy: this meta-topic deals with logistics relating to
immigration, and the subsequent topics are self-explanatory. All laws relate to
overarching immigration policies.
International: Topic six – terrorism – was simple to decipher. Words
in this topic included "isis", "war", "radical", and so on. More crucial to our
project, "refugees", a category of legal immigrant, also co-occurred with language
relating to terrorism. (For the most part, our results found more immigration
discourse related to people entering the United States from the southern border.)
This shows that it was likely for candidates to discuss refugees and terrorism in the
same breath.
America First: Other topics in the analysis related directly to
praising the United States, its people, and its values. "Idealism" included words
like "greatest", "history", and "future", showing how candidates praised the nation’s
accomplishments in a time-specific context. This topic also included “middle-class,”
which consisted of words like “people,” "working", and "families", showing how
candidates collocated these terms in the same instances. Similarly, "nationalism"
collocated words like "americans", "values", and "dream". This data shows the
positive, optimistic language candidates used to refer to the nation in which they
were running for president.
Crime: The "drugs" topic did not specifically contain words that
would normally be associated with immigration. However, it did include words like
"brought", referencing cases in which candidates referred to drugs that had been
brought across the border, indirectly referencing immigration. The topic "gun
control" also seemed to not directly relate to immigration. However, this topic
included words like "terrorists", "san", and "bernardino", relating to the 2015 San
Bernardino attack, in which ISIS claimed responsibility for two shooters with Middle
Eastern heritage. This relates back to the "terrorism" topic, which included the word
"refugees".
Economy: This category also related to the "america first" notion of
valuing nationalism and the work of middle-class Americans. "Economic growth"
included "job" and "cutting", which may refer to jobs that have been filled by
immigrants. Similarly, the "working America" included "support" and "americans", and
the "economy/taxes" topic included "jobs", "minimum", and "wages", which may refer to
immigrant jobs.
Radar Plots The reason for using radar plots is to have a simple comparison for the number of
times candidates use a word. As you can see, there are substantial differences in the
language that democratic and republic candidates use.
The specific figures for the amount of times the republican party used immigration
terms are: 19 for immigrant 15 for undocumented immigrants 17 for
illegal immigrant 31 for refugee and 38 other terms The figures for
the democrats are as follows: 22 for immigrant 13 for undocumented
immigrants 1 for illegal immigrants 19 for refugees and 5 times for
other
These terms are simplified categories for the terms the parties used. Below are
all the specific terms the candidates used in the debates.
Immigrant-immigrant(s), Mexican immigrants, American immigrants, Muslim immigrant
Undocumented-undocu. immigrant, undoc. child, undoc. worker, undoc. people, these
who were undocumented, and undoc. tomato picker. Illegal-Illegal immigrant,
illegal alien, illegals, and aliens. Refugee-refugee, syrian refugee, iraqi
refugee, and Middle Eastern refugee Other-Guest worker, people, our neighbors,
student, those who live in the shadows
The radar plot evidently shows the republican party spoke immigrant terms more
than democrats. In fact, they said these immigrant terms twice the amount of times
democrats did: 120 to 60.
Keywords were significant repeated terms that candidates emphasized.
Interestingly, amnesty is not used at all by democrats; thus, amnesty is a primary issue
in immigration discourse for republicans.
The number for repblicans are:
28 for immigration 107 for border 15 for pathway to citizenship
48 for wall 80 for amnesty 32 for other
The number for democrats are:
49 for immigration 1 for border 10 for pathway to citizenship
10 for wall 0 for amnesty 1 for other
Obviously, the democrats do not generally use the same keywords republicans use.
The only 'other' keyword democrats had was 'asylum' which republicans did not use at
all. There were many variations of immigration:
Border had border security,
border agent, border patrol and border control.
Wall and amnesty had no
variations.
The terms that fell under 'other' were: birthright
citizenship, earned legal status, asylum, e-verify, santuary city, comprehensive reform,
and visa overstay.
Republicans used 310 keywords while democrats used 71.
In other words, republicans used keywords more than four times the amount compared to
democrats.
Democrats used keywords substantially less than republicans even
for the same terms. There is margin for error, but overall, the republicans focused on
different topics when talking about immigration such as a wall, amnesty, and borders.
Democrats are more limited in their discussion of immigration due to their lack of
variety in keywords. Both of the radar plots show democrats using the general
'immigrant' and 'immigration' more than republicans despite their lower word count. It
can be concluded that republicans tend to differentiate immigrants more than democrats,
specifically by legal status and what type of immigrant they are.
Parallel Coordinates
The republicans, naturally red, and democrats, blue, are on a graph which
analyzes the relationship between the proportion of trope types they use out of all the
debates. The republicans dominate trope type and they are mostly focused on
security in proportion to the other trope types.
This parallel coordinate plot looks at the individual candidates. Some
candidates did not show up since they dropped out of running for candidacy and used no
tropes in the debate. The two parallel coordinate plots naturally mirror each other, the
democratic party uses more tropes on values than security than economy. Republicans are
the opposite, they are heavily focused on security and economy rather than values.
Conclusion Based on our research, we were able to determine that generally, Republican
candidates referred to immigration in the 2016 presidential debates more frequently than
their Democratic counterparts. The most popular individual terms included: "illegal
alien", "illegal", "illegal immigrant", "undocumented ____", "dreamer" "asylum-seeker",
"refugee", "migrant", and "guestworker". However, within this discourse, Republican
candidates varied their speech to refer to immigration in pro, anti, and neutral terms,
referencing specific concepts as diverse as "amnesty" and the proverbial "wall". Some
Republican discourse referred to immigrants as "the bad ones" as well as "the good
ones", showing a disparity within the party.