Comparison of State-Level Two-Party Presidential Vote Share: 2012 and 2016

pres_scatter.R

# Library and data import ======================================================
library(tidyverse)
library(scales)
library(ggrepel)
library(Kmisc) # devtools::install_github("sysilviakim/Kmisc")
options(digits = 5, scipen = 999)

# MEDSL Data: U.S. President 1976–2016
# https://doi.org/10.7910/DVN/42MVDX
# This data file contains constituency (state_po-level) returns for elections
# to the U.S. presidency from 1976 to 2016.

url <- 'https://dataverse.harvard.edu/api/access/datafile/3444051?gbrecs=false'
pres_raw <- read.csv(url, sep = "\t")

# Filter data ==================================================================
pres_df <- pres_raw %>%
  filter(year == 2012 | year == 2016) %>%
  filter(candidate != "" & writein == FALSE) %>%
  mutate(
    party = ifelse(party == "democratic-farmer-labor", "democrat", party)
  ) %>%
  filter(party == "democrat" | party == "republican") %>%
  group_by(year, state_po) %>%
  # Since the unit of analysis is candidate-state_po-year, create vote shares
  summarise(
    vote_share = candidatevotes[party == "democrat"] / (
      candidatevotes[party == "democrat"] + 
        candidatevotes[party == "republican"]
    )
  ) %>%
  # Now since we have years 2012 and 2016, but want to make it scatterplot,
  # pivot_wider by each state_po
  pivot_wider(names_from = year, values_from = vote_share)

# Scatterplot ==================================================================
p <- pres_df %>%
  mutate(
    type = case_when(
      `2012` < 0.5 & `2016` < 0.5 ~ 1,
      `2012` < 0.5 & `2016` >= 0.5 ~ 2,
      `2012` >= 0.5 & `2016` < 0.5 ~ 3,
      `2012` >= 0.5 & `2016` >= 0.5 ~ 4
    ),
    type = factor(
      type,
      levels = seq(4),
      labels = c(
        "Consistently Republican", "Republican --> Democrat",
        "Democrat --> Republican", "Consistently Democrat"
      )
    )
  ) %>%
  ggplot(aes(x = `2012`, y = `2016`, colour = type, label = state_po)) + 
  geom_point() + 
  theme_bw() + 
  scale_x_continuous(labels = scales::percent) + 
  scale_y_continuous(labels = scales::percent) + 
  # geom_abline(intercept = 0, slope = 1) + 
  geom_vline(xintercept = .5, linetype = "dotted") + 
  geom_hline(yintercept = .5, linetype = "dotted") +
  # scale_colour_brewer(type = "div", palette = 5) + 
  scale_colour_manual(
    values = c(
      "Consistently Republican" = "#ca0020",
      "Republican --> Democrat" = "#92c5de",
      "Democrat --> Republican" = "#f4a582",
      "Consistently Democrat" = "#0571b0"
    )
  ) + 
  # stat_summary(fun.data = mean_cl_normal) + 
  # geom_smooth(method = 'lm', formula = y ~ x)
  xlab("2012 Two-Party Vote Share") + 
  ylab("2016 Two-Party Vote Share") + 
  theme(legend.title = element_blank())

# Export, with labels or without ===============================================
pdf("pres_scatter_nolabel.pdf", width = 6, height = 4)
Kmisc::pdf_default(p) + 
  theme(legend.position = "botton")
dev.off()

pdf("pres_scatter_labeled.pdf", width = 6, height = 4)
Kmisc::pdf_default(p) + 
  geom_text_repel(show.legend = FALSE) + 
  theme(legend.position = "botton")
dev.off()