The influence of the fake news phenomenon
library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionlibrary(ggplot2)Research Question
I am interested in analyzing if the increasing distrust in the media(media_2019), particularly within the Republican voter base, influences US citizens’ confidence in political institutions and feelings toward the opposing Democratic party.This analysis will include confidence in institutions(inst_court_2019, inst_FBI_2019), and feelings toward Democrats(Democrats_2019).
I will be studying the feelings of voters who have differing degrees of confidence in the media, divided by respondents who agree and disagree in the statement that “you can’t believe much of what you hear from the mainstream media” .
Data Source
This analysis uses the data set from the Democracy Fund Voter Study Group provided by Professor Brett Turner of Queens College.
Data Prep
voter_2019<-read.csv("C:/Users/12055/Documents/CUNY - Undergrad/Spring 2021 - CUNY/Data 333/Assignments/Recoding Variables/Data/Voter Data 2019.csv")
Voter_2019<-voter_2019%>%
mutate(FakeMedia= ifelse(media_2019==8, NA,
ifelse(media_2019<3,"Agree",
ifelse(media_2019>2,"Disagree", NA))),
ConfidenceSupreme= ifelse(inst_court_2019==1,"A great deal",
ifelse(inst_court_2019==2,"Quite a lot",
ifelse(inst_court_2019==3,"Some",
ifelse(inst_court_2019==4,"Very Little", NA)))),
ConfidenceFBI= ifelse(inst_FBI_2019==1,"A great deal",
ifelse(inst_FBI_2019==2,"Quite a lot",
ifelse(inst_FBI_2019==3,"Some",
ifelse(inst_FBI_2019==4,"Very Little", NA)))),
FeelingsDemocrats= ifelse(Democrats_2019<=100, Democrats_2019,
ifelse (Democrats_2019>100,NA, NA
)))%>%
select(FakeMedia, ConfidenceSupreme, ConfidenceFBI, FeelingsDemocrats)%>%
filter(!is.na(FakeMedia), !is.na(ConfidenceSupreme), !is.na(ConfidenceFBI), !is.na(FeelingsDemocrats))
head(Voter_2019)## FakeMedia ConfidenceSupreme ConfidenceFBI FeelingsDemocrats
## 1 Agree Some A great deal 99
## 2 Disagree Very Little Quite a lot 25
## 3 Agree A great deal A great deal 31
## 4 Disagree Some Some 100
## 5 Disagree Some Very Little 53
## 6 Agree Some Some 97colnames(Voter_2019)## [1] "FakeMedia" "ConfidenceSupreme" "ConfidenceFBI"
## [4] "FeelingsDemocrats"dim(Voter_2019)## [1] 5906 4Fake media and confidence in the Supreme Court
Cross tab
Fake media and confidence in the Supreme Court
table(Voter_2019$FakeMedia , Voter_2019$ConfidenceSupreme)%>%
prop.table(1)%>%
round(2)##
## A great deal Quite a lot Some Very Little
## Agree 0.19 0.36 0.35 0.10
## Disagree 0.08 0.24 0.50 0.19Analysis
From the cross tab, we can observe the following:
- Respondents who distrust the media responded with more than double the confidence, or “a great deal” of confidence, in the Supreme Court as respondents who trust the media.
- Respondents who have more trust in the media tended to have less confidence, or “very little” confidence, in the Supreme Court.
Visualization
Fake media and confidence in the Supreme Court
Voter_2019%>%
group_by(FakeMedia,ConfidenceSupreme)%>%
summarize(n=n())%>%
mutate(percent=n/sum(n))%>%
ggplot()+
geom_col(aes(x=FakeMedia, y=percent, fill=ConfidenceSupreme))## `summarise()` has grouped output by 'FakeMedia'. You can override using the `.groups` argument.
Analysis
From the bar chart depicting confidence in the Supreme Court based on distrust in the media, we can observe that greater confidence in the Supreme Court makes up more than half of the responses of voters who distrust the media.
Chi Squared Test
Fake media and confidence in the Supreme Court
chisq.test(Voter_2019$FakeMedia , Voter_2019$ConfidenceSupreme )##
## Pearson's Chi-squared test
##
## data: Voter_2019$FakeMedia and Voter_2019$ConfidenceSupreme
## X-squared = 344.93, df = 3, p-value < 2.2e-16Analysis
There is a statistically significant relationship between the variable FakeMedia and ConfidenceSupreme as the p-value is below 0.05. This implies that belief in fake news media has a significant relationship on the variable reflecting confidence in the Supreme Court.
Fake media and confidence in the FBI
Cross tab
Fake media and confidence in the FBI
table(Voter_2019$FakeMedia , Voter_2019$ConfidenceFBI)%>%
prop.table(1)%>%
round(2)##
## A great deal Quite a lot Some Very Little
## Agree 0.08 0.25 0.44 0.22
## Disagree 0.30 0.40 0.26 0.04Analysis
From the cross tab, we can observe that respondents who distrust the media responded with less than a third of confidence in the FBI as respondents who trust the media.
Visualization
Fake media and confidence in the FBI
Voter_2019%>%
group_by(FakeMedia,ConfidenceFBI)%>%
summarize(n=n())%>%
mutate(percent=n/sum(n))%>%
ggplot()+
geom_col(aes(x=FakeMedia, y=percent, fill=ConfidenceFBI))## `summarise()` has grouped output by 'FakeMedia'. You can override using the `.groups` argument.
Analysis
From the bar chart depicting confidence in the FBI based on distrust in the media, we can observe the following:
Respondents who distrust the media have a markedly lesser degree of confidence in the FBI than respondents who trust the media.
Greater confidence in the FBI amongst respondents who trust the media represents about twice the respective portion of respondents who distrust the media.
Chi Squared Test
Fake media and confidence in the FBI
chisq.test(Voter_2019$FakeMedia , Voter_2019$ConfidenceFBI )##
## Pearson's Chi-squared test
##
## data: Voter_2019$FakeMedia and Voter_2019$ConfidenceFBI
## X-squared = 963.48, df = 3, p-value < 2.2e-16Analysis
There is a statistically significant relationship between the variable FakeMedia and ConfidenceFBI as the p-value is below 0.05. This implies that belief in fake news media has a significant relationship on the variable reflecting confidence in the FBI.
Fake media and feelings toward the opposing political party
Table of Means
Voter_2019%>%
filter(!is.na(FeelingsDemocrats))%>%
group_by(FakeMedia)%>%
summarize(FeelingsDemocrats=mean(FeelingsDemocrats,na.rm=TRUE))## # A tibble: 2 x 2
## FakeMedia FeelingsDemocrats
## <chr> <dbl>
## 1 Agree 25.0
## 2 Disagree 74.0Analysis
From the table of means we can observe that the those who distrust the media responded with a much lower rating of their feelings toward Democrats, reflecting the phenomenon of belief in “fake news” within the Republican party.
Bar chart
Fake media and feelings toward Democrats
Voter_2019%>%
filter(!is.na(FeelingsDemocrats))%>%
group_by(FakeMedia)%>%
summarize(FeelingsDemocrats=mean(FeelingsDemocrats,na.rm=TRUE))%>%
ggplot(aes(x=FakeMedia,y=FeelingsDemocrats,fill=FakeMedia))+
geom_col()
Analysis
The bar chart above is a visualization of the stark disparity in feelings toward Democrats based on distrust of the media. Voters in 2019 who responded with a distrust of the media have markedly lower ratings of feelings toward Democrats.
Visualization
Histogram - Fake Media and Feelings toward Democrats
Voter_2019%>%
select(FakeMedia, FeelingsDemocrats)%>%
filter(FakeMedia %in% c("Agree", "Disagree"))%>%
ggplot()+
geom_histogram(aes(x=FeelingsDemocrats, fill=FakeMedia))+
facet_wrap(~FakeMedia)## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
Analysis
The population distribution of respondents who distrust the media is skewed left based on lower ratings of Democrats. The poulation distribution of respondents who trust the media is skewed right based on greater ratings of Democrats.
Sampling Distributions
Histogram - Fake Media and Feelings toward Democrats
AgreeFM<-Voter_2019%>%
filter(FakeMedia=="Agree")
DisagreeFM<-Voter_2019%>%
filter(FakeMedia=="Disagree")
AgreeFM_Samp_Distro<-replicate(10000, sample(AgreeFM$FeelingsDemocrats, 40)%>%
mean(na.rm=TRUE))%>%
data.frame()%>%
rename("mean"=1)
DisagreeFM_Samp_Distro<-replicate(10000, sample(DisagreeFM$FeelingsDemocrats, 40)%>%
mean(na.rm=TRUE))%>%
data.frame()%>%
rename("mean"=1)
ggplot()+
geom_histogram(data=AgreeFM_Samp_Distro, aes(x=mean), fill="white")+
geom_histogram(data=DisagreeFM_Samp_Distro, aes(x=mean), fill="black")## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
Analysis
Sampling Distributions
The distribution of 2019 voters’ responses based on agreeing that mainstream media cannot be believed versus disagreeing, or trusting the media, and feelings toward Democrats are both distributed normally. Lack of confidence in news media is linked with less positive feelings toward Democrats, as observed with a distribution lower down the x axis.
t-test
Fake Media and Feelings toward Democrats
t.test(Voter_2019$FeelingsDemocrats ~ Voter_2019$FakeMedia)##
## Welch Two Sample t-test
##
## data: Voter_2019$FeelingsDemocrats by Voter_2019$FakeMedia
## t = -70.475, df = 5904, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -50.36119 -47.63525
## sample estimates:
## mean in group Agree mean in group Disagree
## 25.02736 74.02558Analysis
According to the results of the t-test between the variables “FakeMedia” and “FeelingsDemocrats”, there is a significant difference in the mean ratings of Democrats based on distrust of the media. The p value is 2.2x10-6, much smaller than the threshold of confidence (alpha=0.05).
Conclusion
The results of this analysis demonstrate that voters who have less trust in the media are also less confident in FBI and lean to the right on the political spectrum. Voters who have greater trust in the media are less confident in the Supreme Court but more confident in the FBI and lean left on the political spectrum based on feelings toward Democrats.