DATA310-finalproject-political-corruption-analysis

# DATA-3100
# Adefoluke Shemsu

##Setup workspace and console##

# Setting working directory

setwd("~/Documents/Education/Penn/Classes/DATA 310/Week 8")

# Pulling packages that might be needed

library(tidyverse)
library(survey)
library(sandwich)
library(stargazer)

# Pulling data set

load("~/Documents/Education/Penn/Classes/DATA 310/Week 8/ANES20Datafor310.Rdata")

##PART 1: Cleaning data and preparing variables##

# State

anes$state <- anes$V203001

# Biden thermometer

anes$biden.t <- anes$V201151

anes$biden.t[anes$biden.t %in% c(-9,-4,998)] <- NA
summary(anes$biden.t)
table(anes$biden.t)

# Trump thermometer

anes$trump.t <- anes$V201152

anes$trump.t[anes$trump.t %in% c(-9)] <- NA

summary(anes$trump.t)
table(anes$biden.t)

# Respondent political thermometer

anes$poli.t <- anes$V201200 
# Relegating nuanced sentiments to larger political buckets and removing moderates because 
# I want to zoom in on conservative/liberal thoughts
anes$poli.t[anes$poli.t %in% c(-9,-8,99, 4)] <- NA
anes$poli.t[anes$poli.t %in% c(1,2,3)] <- 0
anes$poli.t[anes$poli.t %in% c(5,6,7)] <- 1


##Variables to measure belief in how much a president changes corruption beliefs##

# Trust themometer in Washington to do the right thing

anes$trust.wa <- anes$V201233

anes$trust.wa[anes$trust.wa %in% c(-9,-8)] <- NA
anes$trust.wa[anes$trust.wa %in% c(1,2,3)] <- 0
anes$trust.wa[anes$trust.wa %in% c(4,5)] <- 1


# Government for few wealthy or many citizens

anes$oligopoly.t <- anes$V201234

anes$oligopoly.t[anes$oligopoly.t %in% c(-9,-8)] <- NA
anes$oligopoly.t[anes$oligopoly.t %in% c(1)] <- 0
anes$oligopoly.t[anes$oligopoly.t %in% c(2)] <- 1

# Corruption thermometer since Trump election

anes$corruption.t <- anes$V201380

anes$corruption.t[anes$corruption.t %in% c(-9,-8)] <- NA
anes$corruption.t[anes$corruption.t %in% c(1,3)] <- 0
anes$corruption.t[anes$corruption.t %in% c(2)] <- 1

##Creating weight variables##

# Weight

anes$weight<- anes$V200010a

# Removing unnecessary variables for a cleaner look/easier navigation

anes <- anes[,63:70]

##Begin Report##

# When it comes to corruption, does the president matter? As of late, the prevailing anecdotal sentiment within 
# America's political zeitgeist is that it doesn't matter who sits in that chair because they'll always serve 
# the interest of their corporate and political allies while doing the bare minimum for the citizenry.

# This experiment will attempt to confirm that the public's trust in government doesn't change with President Biden, whether
# or not they like him, using 2020 election data. This hypothesis is built on an observation of the collective frustration
# with our government and elected officials on both sides of the aisle. Though the context in which either side 
# is displeased with our government can vary greatly, our goal is to explore the relationship between political survey 
# respondent's beliefs around government corruption and the prospect of Biden as a president.

# In fewer words, is corruption politically agnostic? If so,  will be reflected in respondents' thoughts on Biden.

##PART 2: Analysis##

# To explore this relationship, I will be using the American National Election Study, an academic study of the 
# American adult populations with a sample size of 8,280 participants.

# My primary variable of interest here is the “Feeling Thermometer”, or 'biden.t', for Joe Biden. 
# It measures how respondents feel toward Biden, with 0 = cold, 50 = neutral, and 100 = hot. 
# A mean of 49.24 for biden.t projects a relative indifference to Biden, which is 10 points higher on average than Donald 
# Trump's thermometer, indicating a slight improvement in how the public feels about him as a presidential candidate. 
# This difference, however, does not necessarily imply that respondents will believe that the government is any less
# corrupt, as this response could come down to overall temperament of both individuals on TV or in debates.

# I chose to include the Trump Thermometer in this analysis because it sets a baseline for the validity of my hypothesis.
# The stark difference in the volume of respondents that felt colder toward Trump than Biden (visualized below) indicates a 
# more positive feeling for Biden over Trump.

summary(anes$biden.t)
plot(density(anes$biden.t, na.rm = T), lwd = 2,
     xlim =c(0,100),
     xlab ="Feeling: Cold(0) to Hot (100)",
     main ="Biden Feeling Thermometer")

summary(anes$trump.t)
plot(density(anes$trump.t,na.rm = T), lwd = 2,
     xlim = c(0,100),
     xlab ="Feeling: Cold(0) to Hot (100)",
     main ="Trump Feeling Thermometer")

table(anes$biden.t > 50)
4036/(4036+4024)

table(anes$trump.t > 50)
3344/(3344+4704)

# 50.07% of respondents had a warm to hot feeling about Biden, as opposed to 41.55% for Trump. 
# This pre-established sentiment lays a foundation on which to avoid negative bias that might impact the number 
# of individuals who believe corruption does not change by supporting Biden. In other words, higher support for Biden
# that also correlates with public sentiments around corruption would support the hypothesis that liking Biden does not
# equate to liking or trusting the government.

# The variables that will help me to understand whether respondents' beliefs around government corruption are impacted by
# a more well-liked presidential candidate are 'trust.wa', 'oligopoly.t', and 'corruption.t'. 

# 'trust.wa' measures how often respondents believe they can trust the government Washington to do the right thing, with 
# trust levels broken down to 0 = 'Over half of the time', 1 = 'Less than half of the time'. Within these figures,
# 'Over half of the time' encompasses 1 = 'Always', 2 = 'Most of the time', 3 = 'About half of the time',
# 4 = 'Some of the time', and 5 = 'Never'. This variable will help me determine whether 
# the number of people who feel warm (> 50 on the Biden Thermometer) toward Biden feel more trusting of 
#the government in comparison to Trump.

# For sake of clarity, I will being isolating this data in a separate set to only include self-identifying conservatives 
# and liberals. This is defined by 'poli.t', which is broken down into binary variables where 0 = 'Extremely liberal', 
# 'Liberal', and 'Slightly liberal' and 1 = 'Slightly conservative', 'Conservative', and 'Extremely Conservative'.
# Moderates, while prudent for political discourse, create ambiguity, as they can sway in either direction, 
# and my main focus is to analyze respondents that fit the primary, binary political philosophies. 

anes2 <- drop_na(anes)

table(anes2$poli.t)

# Once isolated (and once NAs have been removed), the next goal to be to ensure the politcal data is balance. That is to say,
# to ensure both parties are equally represented by our self-identifying sample of respondents. Based on my anaysis,
# we have 2,411 liberals and 2,597 conservatives. Let's see how removing moderates and NAs from the sample changes the Trump
# and Biden thermometers.

plot(density(anes2$biden.t, na.rm = T), lwd = 2,
     xlim =c(0,100),
     xlab ="Feeling: Cold(0) to Hot (100)",
     main ="Biden Feeling Thermometer 2.0")

plot(density(anes2$trump.t,na.rm = T), lwd = 2,
     xlim = c(0,100),
     xlab ="Feeling: Cold(0) to Hot (100)",
     main ="Trump Feeling Thermometer 2.0")

# As seen above, the difference is nominal at best, which allows me to begin investigating the relationship between
# how people feel about Biden and how they feel about the government. That said, now we will begin to review respondents'
# thoughts on corruption. The first variable will be 'trust.wa' below.

barplot(prop.table(table(anes2$trust.wa)),
        xlab="", 
        ylab = "How Often?",
        names.arg = c("Over Half the Time", "Less Than Half the Time"),
        axes = F,
        main = "Will Washington to do the Right Thing?")

# 'oligopoly.t' measures whether respondents believe our government is run by a few wealthy individuals for their own 
# interests, or exists for and by the people. Within this variable, 0 = 'Run by a few big interests' and 1 = 'For the
# benefit of all the people'. This variable will  cement the existence of a relationship between how respondent feels 
# about Biden and whether feeling more strongly for him changes their perspective on who our government serves.

barplot(prop.table(table(anes2$oligopoly.t)),
        xlab = "", 
        ylab = "Big Interests or The People?",
        names.arg = c("A Few Big Interests","Benefit of All"),
        axes = F,
        main = "Who Does Our Government Serve?")

# 'corruption.t' measures whether respondents believe overall government corruption has increased, decreased, or stay the 
# same since Trump was elected. Within it, 0 = 'Increased or Not Change, and 1 = 'Decreased'. I chose to group Increased
# and Not Changed because they collectively reflect a more a pessimistic view of government.
# This variable in particular should help us to better understand whether feelings about Biden are consistent 
# with the hypothesis that people still distrust government because the presence of a warmer feeling
# toward Biden and a belief that corruption has decreased since Trumps election, for example, proves the null hypothesis.

barplot(prop.table(table(anes2$corruption.t)),
        xlab = "", 
        ylab = "Corruption Has:",
        names.arg = c("Increased or Stayed Same", "Decreased"),
        axes = F,
        main = "Gov Corruption Since Trump's Election")

# Below, I have modeled a series of regression analyses of the relationship between respondent's feeling about Biden
# and their feelings about government corruption, as defined by the aforementioned variables.

# Biden Thermometer and Respondent Trust in DC to do the Right Thing:

biden.dctrust <- lm(biden.t ~ trust.wa, data = anes2)
summary(biden.dctrust)

# First, we can see that a p-value that is > .05 (.692), which indicates that there isn't a significant influence or 
# relationship between the feelings for Biden and a belief in weather the government will generally do the right thing. 
# This fails to reject the null hypothesis.

# Biden Thermometer and Respondent Belief that Government Serves Interests Over People:

biden.oli <- lm(biden.t ~ oligopoly.t, data = anes2)
summary(biden.oli)

#Comparison to Trump for context:

trump.oli <- lm(trump.t ~ oligopoly.t, data = anes2)
summary(trump.oli)

# Next is the belief in a significant relationship between the Biden Thermometer and respondents' thoughts on whether
# the government serves Big Interests over the benefit of all. In this analysis, the p-value of 3.03e-11 tells
# us that there is a very significant influence that rejects the null hypothesis. The intercept of 47.7 represents the 
# mean feeling of a respondent whose answer to 'oligopoly.t' was 0. Because 0 = 'Run by a few big interests', this
# leads me to the tentative conclusion that conservatives and liberals who felt relatively cool toward Biden (while still
# feeling even colder toward Trump with an intercept of 39.84) also believe that the government is run by a few big interests
# and does not serve the good of the people.

# Biden Thermometer and Respondent Belief that Corruption has Changed since Trump's election:

biden.corrupt <- lm(biden.t ~ corruption.t, data = anes2)
summary(biden.corrupt)

#Comparison to Trump for context:

trump.corrupt <- lm(trump.t ~ corruption.t, data = anes2)
summary(trump.corrupt)

# Last is the relationship between the belief in how Trump's presidency propagated corruption and how respondents
# feel about Biden. The p-value of 2.2e-16 demonstrates another significant influence that rejects the null hypothesis. 
# The intercept of 52.58 represents the mean feeling of a respondent whose answer to 'corruption.t' was 0, where 0 = 
# 'Increase or Stayed the Same'. This leads me to the believe that conservatives and liberals who felt positively toward
# Biden also feel as if Trump's 2016 presidency led had no impact on or exacerbated government corruption.

# Conclusion and Takeaways

# Overall, my hypothesis that self-identifying, binary (conservative vs liberal) political respondents' responses would
# reflect an apathetic, pessimistic view of government was not proven. Conversely, this data tells a more logical,
# common-sense story.

# The story that I've gleaned from this data analysis is that a respondent's beliefs in a corruptible or ineffective  government
# is also tied to a negative feeling both toward Biden and Trump, regardless of whether they're a conservative or liberal.
# While there is a significant correlation between the two, my initial hypothesis drew the wrong conclusion, as it appears
# that there is a still correlation with how people feel about Biden specifically. 

# For example, in the 'biden.corrupt' regression that analyzes the relationship between 'bidet.t' and 'corrupt.t', the 
# average respondent feeling toward Biden was positive (50 > on thermometer). Given that these same respondents generally
# felt drastically more negatively toward Trump, this implies that that both liberals and conservatives both do care
# about who sits in that presidential seat. Comparing that same 'corrupt.t' data against 'trump.t' which produced an intercept
# of 34.28 supports this notion.

# Overall, it appears that I would have be easier to reject a null hypothesis if my broadened my hypothesis. Rather than 
# focus on whether respondent's didn't care, I could instead have hypothesized about the relationship between a respondent's
# feeling toward Biden in comparison to their feelings about Trump, then use that information to understand how Trump's presidency 
# might have impact a respondent's perspective on corruption and servant leadership at a federal level. 

# Recalling my initial statements about a common-sense story, this is a summary of my findings:
# - When it comes to a respondent's trust in Washington to do the right thing, there is no correlation with how they 
# feel about Biden.
# - When it comes to a respondent's belief that our government is run by large corporate interests, the correlation 
# was actually found in those individuals who are displeased with both candidates, given that the mean temperature for
# both candidates was < 50, which equates a cool to cold feeling.
# - When it comes to the belief that Trump's presidency increased, decreased, or maintained levels of government corruption,
# this was the only variable where the mean was disproportionately positive toward Biden vs Trump, indicating a strong
# correlation between a respondent's belief about Trump's previous performance and their desire to see a new White House
# under Biden. This also may demonstrate a built-in bias in respondents, given the warmer feeling toward Biden.