Figures

In [1]:


set.seed(1234)
library(tidyverse)

── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
✔ dplyr     1.1.4     ✔ readr     2.1.5
✔ forcats   1.0.0     ✔ stringr   1.5.1
✔ ggplot2   3.5.2     ✔ tibble    3.3.0
✔ lubridate 1.9.4     ✔ tidyr     1.3.1
✔ purrr     1.0.4     
── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag()    masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors


library(patchwork)
library(pewmethods)
library(ggstats)

In [2]:


# Load data
load(here::here('data', "research-paper-data-20241029.Rdata"))


# shorten and 'unify' factor levels
# Note: this removes variable labels
data <- data |>
  # use `pewmethods` to preserve order of factor levels
  mutate(across(c(q19, q20, q21, q22, q23, q24,
                  q30, q31, q32, q33, q34, q35),
                ~pewmethods::fct_case_when(
    .x %in% c("Not at all confident",
              "Not at all committed",
              "Not at all concerned") ~ "Not at all confident",
    .x %in% c("Not too confident",
              "Not too committed",
              "Not too concerned",
              "Somewhat unconcerned") ~ "Not too confident",
    .x %in% c("Somewhat confident",
              "Somewhat committed",
              "Somewhat concerned") ~ "Somewhat confident",
    .x %in% c("Very confident",
              "Very committed",
              "Very concerned") ~ "Very confident",
    TRUE ~ NA
  ))) |> 
  # unify factor levels
  mutate(across(c(q25, q36), ~ pewmethods::fct_case_when(
    .x == "Very concerned" ~ "Very concerned",
    .x == "Somewhat concerned" ~ "Somewhat concerned",
    .x %in% c("Not too concerned", "Somewhat unconcerned") ~ "Not too concerned",
    .x == "Not at all concerned" ~ "Not at all concerned",
    TRUE ~ NA))) |>
  # reverse factor levels
  mutate(across(c(q25, q36), ~forcats::fct_rev(.)))

In [3]:


# set ggplot theme that will be used for every plot
# theme_get()
old <- ggplot2::theme_set(ggplot2::theme_bw())  # capture current theme

# set/update new theme
ggplot2::theme_set(
  new = ggplot2::theme_bw(
    base_family = 'TeX Gyre Pagella')+ # set font of ggplots
    theme(legend.position = 'bottom')  # set legend position for all plots
)

In [4]:


# custom function used to create Likert Plots

# To avoid excessive copy/paste, I made a function to create Likert plots
# it is mostly a wrapper around `ggstats::gglikert`, but with a specific set up
# particular colors, theme, etc. 
likert_plot <- function(data, x, ...,  
                        symmetric = FALSE,
                        variable_labels = NULL,
                        label_size = 2.5,
                        vline = FALSE,
                        title = waiver(), 
                        subtitle = waiver(),
                        caption = waiver(),
                        xlab = waiver()){
  
  p <- ggstats::gglikert(
    data = data, include = {{ x }}, ...,
    variable_labels = variable_labels,
    symmetric = symmetric)+
    
    # customize color
    ggplot2::scale_fill_grey(start = 0.5, end = 0.1)+
    # custom theme
    theme_bw(base_family = "TeX Gyre Pagella")+
    theme(
      legend.position = 'bottom',  # place legend on bottom
      axis.title.y.left = element_blank(), # blank out y-axis label
      axis.text.x = element_blank(), # remove percentage text along x-axis
      strip.text.y.right = element_text(angle = 0) # make facet label horizontal
        )
    
    if (vline==TRUE) {
    p <- p + geom_vline(xintercept = 0, color = 'black', linewidth = 1.2)
  } else {
      p
  }
  
 
  p <- p + ggplot2::labs(..., title = title, subtitle=subtitle, caption=caption, x = xlab)
  p

}

In [5]:


# function for creating barplots
custom_barplot <- function(data, x, group, title = NULL, subtitle = NULL,
                           caption = NULL, legend_fill = NULL,
                           xlab = NULL){
  b <- data |> 
    group_by({{ group }}, {{ x }}) |> 
    count() |> 
    drop_na() |> 
    group_by({{ group }}) |> 
  mutate(prop = round(n/sum(n), digits = 3),
         pct = prop*100,
         res = str_c(pct,'% (', n, ')', sep = "")) |> 
  ggplot(aes(x = {{ x }}, y = pct, fill = {{ group }}))+
  geom_bar(position = 'dodge', stat = 'identity')+
  geom_text(aes(label = res), position = position_dodge(1.0), size = 2.5, vjust = -0.5)+
  scale_fill_grey(start = 0.5, end = 0.1)
  
  b <- b + ggplot2::labs(
    title = title,
    subtitle = subtitle,
    caption = stringr::str_wrap(caption, width = 99),
    fill = legend_fill,
    y = "Percentage",
    x = stringr::str_wrap(xlab, width = 85))
  b
}

In [6]:


# assign items to particular vectors to avoid copy/pasting
trust.az.items <- data |> select(q19, q20, q21, q22, q24) |> colnames() |> dput()
trust.lcl.items <- data |> select(q30, q31, q32, q33, q35)|> colnames() |> dput()
distrust.az.items <- data |> select(q28_1:q28_5)|> colnames() |> dput()
distrust.lcl.items <- data |> select(q40_1:q40_5)|> colnames() |> dput()

# variable labels for az and local area items, respectively
trust.az.varlabels <- c(
  q19 = "Votes counted as intended",
  q20 = "Election officials will do good job",
  q21 = "Election workers will be committed",
  q22 = "Voting process will be fair",
  q24 = "Election technology will be secure"
)

distrust.az.varlabels <- c(
  q28_1 = "Voter fraud will occur",
  q28_2 = "Many votes won't be counted",
  q28_3 = "Many voters will be turned away",
  q28_4 = "Foreign Interference",
  q28_5 = "Election officials will discourage voting"
)

trust.lcl.varlabels <- c(
  q30 = "Votes counted as intended",
  q31 = "Election officials will do good job",
  q32 = "Election workers will be committed",
  q33 = "Voting process will be fair",
  q35 = "Election technology will be secure"
)

distrust.lcl.varlabels <- c(
  q40_1 = "Voter fraud will occur",
  q40_2 = "Many Votes won't be counted",
  q40_3 = "Many voters will be turned away",
  q40_4 = "Foreign Interference",
  q40_5 = "Election officials will discourage voting"
)

Trust in Elections across sample

In [7]:


# plot of trust in elections  

# plot of trust in AZ elections 
trust.likertplot.az <- data |> likert_plot(
  x = all_of(trust.az.items),
  variable_labels = trust.az.varlabels,
  title = "Trust in Maricopa County, AZ, Elections"
)

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.


trust.likertplot.az

In [8]:


# plot of trust in Local Area elections 
trust.likertplot.lcl <- likert_plot(
  data = data, 
  x = all_of(trust.lcl.items),
  variable_labels = trust.lcl.varlabels,
  title = "Trust in Local Area Elections")

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.


trust.likertplot.lcl

In [9]:


# Instead of two separate plots, arrange into one plot while remaining separated
p1.trust <- trust.likertplot.az / trust.likertplot.lcl+
  patchwork::plot_annotation(caption = "Missing values omitted")+
  patchwork::plot_layout(guides = "collect") &
  theme(legend.position = "bottom")

p1.trust

Distrust in elections across sample

In [10]:


# responses to distrust in AZ election items, overall sample
distrust.likert.az <- likert_plot(
  data, 
  x = all_of(distrust.az.items), 
  variable_labels = distrust.az.varlabels,
  title = "Distrust in Maricopa County, AZ Elections", 
  subtitle = str_wrap(
    "How likely do you think any or all of the following will happen during this year´s elections in Maricopa County, AZ?",width = 95))

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.


distrust.likert.az

Distrust in Maricopa County, AZ Elections

In [11]:


# responses to distrust in local election items, overall sample
distrust.likert.lcl <- likert_plot(
  data,
  x = all_of(distrust.lcl.items),
  variable_labels = distrust.lcl.varlabels,
  title = "Distrust in Local Area Elections",
  subtitle = str_wrap(
    "How likely do you think any or all of the following will happen during this year´s elections in your local area?",
    width = 95
  )
)

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.


distrust.likert.lcl

In [12]:


p2.distrust <- distrust.likert.az/distrust.likert.lcl+
  patchwork::plot_annotation(caption = "Missing values omitted")+
  patchwork::plot_layout(guides = "collect") &
  theme(legend.position = "bottom")

p2.distrust

Concerns for Violence and Confidence in Voter Safety Across Sample

In [13]:


# plot of safety concerns and confidence :::::::::::::::::::::::::::: 

safety.concerns.likert <- likert_plot(
  data, x = c(q25, q36), 
  y_label_wrap = 15, 
  variable_labels = c(
    q25 = 'Thinking about Maricopa County, AZ',
    q36 = "Thinking about Local Area"), 
  title = "Concern for Voter Intimidation, Threats, or Violence",
  subtitle = str_wrap(
    "How concerned should voters feel about potential violence, threats of violence, or intimidation while voting in person at their local polling place?", width = 99))

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.


safety.concerns.likert

In [14]:


# q26 In-person voter safety at polling sites
safety.conf.likert <- likert_plot(
  data,
  x = c(q26, q37),
  y_label_wrap = 15,
  variable_labels = c(q26 = "In Maricopa County, AZ", q37 = "In your local area"),
  title = str_wrap("Confidence for In-person voter safety at Polling Sites", width = 95),
  subtitle = str_wrap(
    "How confident, if at all, are you that in person polling places will be safe places for voters to cast their ballots during the upcoming elections in November?",
    width = 95
  )
)

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.


safety.conf.likert

In [15]:


p3.safety <- safety.concerns.likert/safety.conf.likert+
  patchwork::plot_annotation(caption = "Missing values omitted")+
  patchwork::plot_layout(guides = "keep") &
  theme(legend.position = "bottom")

Trust and Distrust by Education, Gender, and Partisanship

In [16]:


# plots of trust in AZ elections by educational attainment :::::::::: 
trust.az.educ.likert <- data |>   
  filter(!is.na(educ_4cat)) |> 
  mutate(
    educ_3cat = pewmethods::fct_case_when(
      educ_4cat %in% c("H.S. or less") ~ "H.S. or less",
      educ_4cat %in% c("Some college no degree") ~ "Some college no Degree",
      educ_4cat %in% c("Postgraduate degree", "College degree") ~ "College Graduate")
    ) |>  
  likert_plot(x = all_of(trust.az.items), y = 'educ_3cat', facet_rows = vars(.question),
              facet_label_wrap = 15,
              variable_labels = trust.az.varlabels,
              subtitle = "Maricopa County, AZ")

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [17]:


# plots of trust in Local area elections by educational attainment :: 
trust.lcl.educ.likert <- data |>  
  filter(!is.na(educ_4cat)) |>  
  mutate(
    educ_3cat = pewmethods::fct_case_when(
      educ_4cat %in% c("H.S. or less") ~ "H.S. or less",
      educ_4cat %in% c("Some college no degree") ~ "Some college no Degree",
      educ_4cat %in% c("Postgraduate degree", "College degree") ~ "College Graduate")
    ) |>  
  likert_plot(
    x = all_of(trust.lcl.items), y = 'educ_3cat', facet_rows = vars(.question),
    facet_label_wrap = 15, variable_labels = trust.lcl.varlabels,
    subtitle = "Local Area"
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [18]:


p4.trust.az.educ <- trust.az.educ.likert + trust.lcl.educ.likert + 
    patchwork::plot_layout(guides = "collect") &
    theme(legend.position = "bottom")

In [19]:


# plot of distrust in AZ elections by educational attainment :::::::: 
distrust.az.educ.likert <- data |> 
  filter(!is.na(educ_4cat)) |> 
  mutate(
    educ_3cat = pewmethods::fct_case_when(
      educ_4cat %in% c("H.S. or less") ~ "H.S. or less",
      educ_4cat %in% c("Some college no degree") ~ "Some college no Degree",
      educ_4cat %in% c("Postgraduate degree", "College degree") ~ "College Graduate")
    ) |>  
  likert_plot(x = all_of(distrust.az.items), y = 'educ_3cat', facet_rows=vars(.question),
              facet_label_wrap = 15, variable_labels = distrust.az.varlabels,
              subtitle = "Maricopa County, AZ")

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [20]:


# plot of distrust items by educational attainment
distrust.lcl.educ.likert <- data |> 
  filter(!is.na(educ_4cat)) |> 
  mutate(
    educ_3cat = pewmethods::fct_case_when(
      educ_4cat %in% c("H.S. or less") ~ "H.S. or less",
      educ_4cat %in% c("Some college no degree") ~ "Some college no Degree",
      educ_4cat %in% c("Postgraduate degree", "College degree") ~ "College Graduate")
    ) |>   
  likert_plot(
    x = all_of(distrust.lcl.items), y = 'educ_3cat', facet_rows = vars(.question),
    facet_label_wrap = 15, variable_labels = distrust.lcl.varlabels,
    subtitle = "Local Area"
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [21]:


p5.distrust.az.educ <- distrust.az.educ.likert + distrust.lcl.educ.likert +
    patchwork::plot_layout(guides = "collect", axes = "collect") & 
    theme(legend.position = "bottom")


p5.distrust.az.educ

In [22]:


# plots of trust and distrust in AZ elections by gender ::::::::::::: 
trust.az.gender.likert <- data |>  
  filter(!is.na(gender_3cat), gender_3cat != 'Other/Refused') |> 
  likert_plot(
    x = all_of(trust.az.items), y = 'gender_3cat', facet_rows = vars(.question),
    facet_label_wrap = 15, variable_labels = trust.az.varlabels,
    subtitle = "Maricopa County, AZ Elections"
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [23]:


# plots of distrust in AZ elections by gender ::::::::::::: 
distrust.az.gender.likert <- data |> 
  filter(!is.na(gender_3cat), gender_3cat != 'Other/Refused') |>
  likert_plot(
    x = all_of(distrust.az.items), y = 'gender_3cat', facet_rows = vars(.question),
    facet_label_wrap = 15, variable_labels = distrust.az.varlabels,
    subtitle = "Maricopa County, AZ Elections"
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [24]:


p6.trust.az.gender <- (trust.az.gender.likert + distrust.az.gender.likert) +
  patchwork::plot_annotation(
    caption = str_wrap(
      "Gender Non-binary (n=7) and those who preferred not to say (n=9) response categories omitted for lack of observations. Missing values omitted (n=15)",
      width = 80
    ),
    theme = theme(text = element_text(family = "TeX Gyre Pagella"))
  ) +
  patchwork::plot_layout(
    ncol = 1, 
    nrow = 2,
    guides = 'keep', axes = "collect") &
  theme(legend.position = "bottom")

p6.trust.az.gender

Trust in Maricopa County, AZ Elections by Gender

In [25]:


# plot of trust in local area elections by gender
trust.lcl.gender.likert <- data |>  
  filter(!is.na(gender_3cat), gender_3cat != 'Other/Refused') |> 
  likert_plot(
    x = all_of(trust.lcl.items), y = 'gender_3cat',
    facet_rows = vars(.question),
    facet_label_wrap = 15, variable_labels = trust.lcl.varlabels,
    subtitle = "Local Area Elections"
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [26]:


# plot of distrust in local area elections by gender
distrust.lcl.gender.likert <- data |> 
  filter(!is.na(gender_3cat), gender_3cat != 'Other/Refused') |> 
  likert_plot(
    x = all_of(distrust.lcl.items), y = 'gender_3cat',
    facet_rows = vars(.question),
    facet_label_wrap = 15, variable_labels = distrust.lcl.varlabels,
    subtitle = "Local Area Elections"
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [27]:


p7.trust.lcl.gender <- trust.lcl.gender.likert/distrust.lcl.gender.likert+
  patchwork::plot_annotation(
    caption = str_wrap("Gender Non-binary (n=7) and those who preferred not to say (n=9) response categories omitted for lack of observations. Missing values omitted (n=15)",
                       width = 80), 
    theme = theme(text = element_text(family = "TeX Gyre Pagella"))) +
  patchwork::plot_layout(guides = 'keep', axes = "collect") &
  theme(legend.position = "bottom")

p7.trust.lcl.gender

In [28]:


# plot of trust in AZ elections by partisanship ::::::: 
trust.az.party.likert <- data |>  
  filter(!is.na(partyid_3cat)) |> 
  likert_plot(
    x = all_of(trust.az.items), y = 'partyid_3cat',
    facet_rows = vars(.question), 
    facet_label_wrap = 15,
    variable_labels = trust.az.varlabels,
    subtitle = "Maricopa County, AZ Elections"
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [29]:


# plot of distrust in AZ elections by partisanship ::::::: 
distrust.az.party.likert <- data |> 
  filter(!is.na(partyid_3cat)) |> 
  likert_plot(
    x = all_of(distrust.az.items), y = 'partyid_3cat',
    facet_rows = vars(.question), 
    facet_label_wrap = 15,
    variable_labels = distrust.az.varlabels,
    subtitle = "Maricopa County, AZ Elections"
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [30]:


p8.trust.az.party <- trust.az.party.likert/distrust.az.party.likert +
  patchwork::plot_layout(guides = 'keep', axes = "collect") &
  theme(legend.position = "bottom")

p8.trust.az.party

In [31]:


# plot of trust and distrust in local area elections by partisanship  
trust.lcl.party.likert <- data |>  
  filter(!is.na(partyid_3cat)) |> 
  likert_plot(
    x = all_of(trust.lcl.items), y = 'partyid_3cat',
    facet_rows = vars(.question), 
    facet_label_wrap = 15,
    variable_labels = trust.lcl.varlabels,
    subtitle = "Local Area Elections"
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [32]:


# plot of distrust in local area elections by partisanship  
distrust.lcl.party.likert <- data |> 
  filter(!is.na(partyid_3cat)) |>  
  likert_plot(
    x = all_of(distrust.lcl.items), y = 'partyid_3cat',
    facet_rows = vars(.question), 
    facet_label_wrap = 15,
    variable_labels = distrust.lcl.varlabels,
    subtitle = "Local Area Elections"
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [33]:


p9.trust.lcl.party <- trust.lcl.party.likert/distrust.lcl.party.likert+
  patchwork::plot_layout(guides = 'keep', axes = "collect") &
  theme(legend.position = "bottom")

p9.trust.lcl.party

Trust and Distrust by Treatment Condition

In [34]:


# plot of trust in AZ elections by treatment condition
trust.likert.treat.az <- data |>
  likert_plot(x = all_of(trust.az.items), y = 'group', 
    facet_rows = vars(.question),
    subtitle = "Maricopa County, Arizona")+
  theme(strip.text.y.right = element_blank())

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [35]:


# plot of trust in local elections by treatment condition
trust.likert.treat.lcl <- data |>
  likert_plot(x = all_of(trust.lcl.items), y = 'group', 
    facet_rows = vars(.question),
    facet_label_wrap = 10,
    variable_labels = trust.lcl.varlabels, 
    subtitle = "Local Area")+
  theme(strip.text.y.right = element_blank()) +
  theme(
    axis.text.y.left = element_blank(), # remove y-axis text on 2nd plot
    strip.text.y.right = element_text(angle = 0, size = 8)
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [36]:


p10.trust.treat <- trust.likert.treat.az + trust.likert.treat.lcl+
  patchwork::plot_layout(guides = "collect") &
  theme(legend.position = "bottom")

p10.trust.treat

Trust in Election Administration by Experiment Condition

In [37]:


# distrust in AZ elections by experimental condition
distrust.az.treat.likert <- data |>
  likert_plot(x = all_of(distrust.az.items), y = 'group', 
    facet_rows = vars(.question),
    subtitle = "Maricopa County, AZ")+
  theme(strip.text.y.right = element_blank())

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [38]:


# distrust in local area elections by treatment condition
distrust.lcl.treat.likert <- data |>
  likert_plot(x = all_of(distrust.lcl.items), y = 'group', 
    facet_rows = vars(.question),
    facet_label_wrap = 10,
    variable_labels = distrust.lcl.varlabels,
    subtitle = "Local Area")+
  theme(strip.text.y.right = element_blank()) +
  theme(
    axis.text.y.left = element_blank(), # remove y-axis text on 2nd plot
    strip.text.y.right = element_text(angle = 0, size = 8)
  )

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [39]:


p11.distrust.treat <- distrust.az.treat.likert + distrust.lcl.treat.likert+
  patchwork::plot_layout(guides = 'collect')

p11.distrust.treat

Disrust in Election Administration by Experiment Condition

trust and distrust by partisanship and treatment

In [40]:


# plots of trust and distrust by partisanship and treatment ::::::::: 

# plot of trust in AZ elections by experiment condition and party id
trust.az.treat.partyid <- data |> 
  filter(!is.na(partyid_3cat)) |>
  likert_plot(
    x = all_of(trust.az.items),
    y = 'partyid_3cat',
    facet_rows = vars(.question),
    facet_cols = vars(group),
    facet_label_wrap = 15,
    variable_labels = trust.az.varlabels,
    subtitle = "Maricopa County, AZ")

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [41]:


# plot of trust in local elections by experiment condition and party id
trust.lcl.treat.partyid <- data |> 
  filter(!is.na(partyid_3cat)) |>
  likert_plot(
    x = all_of(trust.lcl.items),
    y = 'partyid_3cat',
    facet_rows = vars(.question),
    facet_cols = vars(group),
    facet_label_wrap = 15,
    variable_labels = trust.lcl.varlabels,
    subtitle = "Local Area")

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [42]:


p12.trust.treat.partyid <- trust.az.treat.partyid / trust.lcl.treat.partyid+
  patchwork::plot_layout(guides = 'collect')

p12.trust.treat.partyid

Trust in Elections by Treatment and Partisanship

Trust and Distrust by Legitimacy beliefs and Treatment Condition

In [43]:


# plots of trust and distrust by legitimacy and treatment ::::::::::: 

# trust in AZ elections by experiment conditions among those who believe the 2020
# election results were illegitimate
trust.az.treat.legit <- data |> 
  filter(!is.na(partyid_3cat), !is.na(q7)) |>
  filter(q7 == "Not legitimate") |> 
  likert_plot(
    x = all_of(trust.az.items), 
    y = 'group',
    facet_rows = vars(.question),
    facet_cols = vars(q7),
    facet_label_wrap = 15,
    variable_labels = trust.az.varlabels,
    subtitle = str_wrap("Maricopa County, AZ",
      width = 95) 
  )+
  theme(text = element_text(family = 'TeX Gyre Pagella'),
        strip.text.y.right = element_blank())

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [44]:


# trust in local elections by experiment conditions among those who believe the
# 2020 election results were illegitimate
trust.lcl.treat.legit <- data |> 
  filter(!is.na(partyid_3cat), !is.na(q7)) |>
  filter(q7 == "Not legitimate") |> 
  likert_plot(
    x = all_of(trust.lcl.items), 
    y = 'group',
    facet_rows = vars(.question),
    facet_cols = vars(q7),
    facet_label_wrap = 15,
    variable_labels = trust.lcl.varlabels,
    subtitle = str_wrap("Local Area",
      width = 95) 
  )+
  theme(
    text = element_text(family = 'TeX Gyre Pagella'),
    axis.text.y.left = element_blank())

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [45]:


p13.trust.treat.legit <- trust.az.treat.legit + trust.lcl.treat.legit+
  patchwork::plot_layout(guides = 'collect')

p13.trust.treat.legit

Trust in elections by treatment among those who refute the 2020 election results

In [46]:


# distrust in AZ elections by experiment conditions among those who believe the
# 2020 election results were illegitimate
distrust.az.treat.legit <- data |> 
  filter(!is.na(partyid_3cat), !is.na(q7)) |>
  filter(q7 == "Not legitimate") |> 
  likert_plot(
    x = all_of(distrust.az.items), 
    y = 'group',
    facet_rows = vars(.question),
    facet_cols = vars(q7),
    facet_label_wrap = 15,
    variable_labels = distrust.az.varlabels,
    subtitle = str_wrap("Maricopa County, AZ",
      width = 95) 
  )+
  theme(
    text = element_text(family = 'TeX Gyre Pagella'),
    strip.text.y.right = element_blank())

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [47]:


# distrust in local elections by experiment conditions among those who believe
# the 2020 election results were illegitimate
distrust.lcl.treat.legit <- data |> 
  filter(!is.na(partyid_3cat), !is.na(q7)) |>
  filter(q7 == "Not legitimate") |> 
  likert_plot(
    x = all_of(distrust.lcl.items), 
    y = 'group',
    facet_rows = vars(.question),
    facet_cols = vars(q7),
    facet_label_wrap = 15,
    variable_labels = distrust.lcl.varlabels,
    subtitle = str_wrap("Local Area",
      width = 95) 
  )+
  theme(
    text = element_text(family = 'TeX Gyre Pagella'),
    axis.text.y.left = element_blank())

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [48]:


p14.distrust.treat.legit <- distrust.az.treat.legit + distrust.lcl.treat.legit+
  patchwork::plot_layout(guides = 'collect')

p14.distrust.treat.legit

Distrust in elections by treatment among those who refute the 2020 election results

Concerns for Violence and Confidence in Voter Safety by Treatment

In [49]:


# plot for safety concerns by treatment ::::::::::::::::::::::::::::: 

# Because q25 measures levels of concern/worry, it cannot be included in the
# same Likert plot as q26 as that measures reported confidence that voters will
# be safe to vote in-person

# safety concerns by treatment
safety.concern.treat.likert <- data |>
  # Need to slightly modify text of factor levels
  mutate(across(
    c(q25, q36),
    ~ pewmethods::fct_case_when(
      .x == "Very concerned" ~ "Very concerned",
      .x == "Somewhat concerned" ~ "Somewhat concerned",
      .x %in% c("Not too concerned", "Somewhat unconcerned") ~ "Not too concerned",
      .x == "Not at all concerned" ~ "Not at all concerned",
      TRUE ~ NA
    )
  )) |>
  likert_plot(
    x = c(q25, q36),
    y = 'group',
    facet_rows = vars(.question),
    facet_label_wrap = 15,
    variable_labels = c(q25 = 'Thinking about Maricopa County, AZ', q36 = "Thinking about Local Area"),
    subtitle = str_wrap(
      "How concerned should voters feel about potential violence, threats of violence, or intimidation while voting in person at their local polling place?",
      width = 99
    )
  )+
  theme(
    text = element_text(family = 'TeX Gyre Pagella'), 
    axis.title.y.left = element_blank())

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [50]:


# q26 In-person voter safety at election sites by treatment
safety.conf.treat.likert <- data |>
  likert_plot(
    x = c(q26, q37), 
    y = 'group',
    facet_label_wrap = 15,
    facet_rows = vars(.question),
    variable_labels = c(
      q26 = "In Maricopa County, AZ",
      q37 = "In your local area"
    ),
    subtitle = str_wrap("How confident, if at all, are you that in person polling places will be safe places for voters to cast their ballots during the upcoming elections in November?", width = 95)
  )+
  theme(
    text = element_text(family = 'TeX Gyre Pagella'), 
    axis.title.y.left = element_blank())

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

In [51]:


p15.safety.treat <- safety.concern.treat.likert/safety.conf.treat.likert+
  patchwork::plot_layout(guides = "keep")+
  patchwork::plot_annotation(
    title = str_wrap("Safety Concerns by Experiment Condition", width = 80), 
    tag_levels = "a")

p15.safety.treat

An alternative

In [52]:

#|   - "Confidence for In-person voter safety at Election Sites"
#|   - "Concern for Voter Intimidation, Threats, or Violence"

# safety concerns by treatment
safety.concern.treat.likert <- data |>
  # Need to slightly modify text of factor levels
  mutate(across(
    c(q25, q36),
    ~ pewmethods::fct_case_when(
      .x == "Very concerned" ~ "Very concerned",
      .x == "Somewhat concerned" ~ "Somewhat concerned",
      .x %in% c("Not too concerned", "Somewhat unconcerned") ~ "Not too concerned",
      .x == "Not at all concerned" ~ "Not at all concerned",
      TRUE ~ NA
    )
  )) |>
  likert_plot(
    x = c(q25, q36),
    y = 'group',
    facet_rows = vars(.question),
    facet_label_wrap = 15,
    variable_labels = c(q25 = 'Thinking about Maricopa County, AZ', q36 = "Thinking about Local Area"),
    subtitle = str_wrap(
      "How concerned should voters feel about potential violence, threats of violence, or intimidation while voting in person at their local polling place?",
      width = 99
    )
  )+
  theme(
    text = element_text(family = 'TeX Gyre Pagella'), 
    axis.title.y.left = element_blank())

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

#|   - "Confidence for In-person voter safety at Election Sites"
#|   - "Concern for Voter Intimidation, Threats, or Violence"

# In-person voter safety at election sites by treatment
safety.conf.treat.likert <- data |>
  likert_plot(
    x = c(q26, q37), 
    y = 'group',
    facet_label_wrap = 15,
    facet_rows = vars(.question),
    variable_labels = c(
      q26 = "In Maricopa County, AZ",
      q37 = "In your local area"
    ),
    subtitle = str_wrap("How confident, if at all, are you that in person polling places will be safe places for voters to cast their ballots during the upcoming elections in November?", width = 95)
  )+
  theme(
    text = element_text(family = 'TeX Gyre Pagella'), 
    axis.title.y.left = element_blank())

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.

#|   - "Confidence for In-person voter safety at Election Sites"
#|   - "Concern for Voter Intimidation, Threats, or Violence"

# fig a
safety.concern.treat.likert

# fig b
safety.conf.treat.likert

Distribution of Confidence in Elections

In [53]:


ggplot(data) +
  geom_histogram(
    aes(x = conf.az.rescaled , y = after_stat(density)),
    bins = 15,
    alpha = 0.2,
    color = 1,
    fill = "red"
  ) +
  geom_histogram(
    aes(x = conf.lcl.rescaled , y = after_stat(density)),
    bins = 15,
    alpha = 0.2,
    color = 1,
    fill = "blue"
  ) +
  geom_segment(
    x = round(mean(data$conf.az.rescaled), 2),
    y = 0,
    yend = 0.35,
    color = "firebrick",
    lwd = 1.4,
    linetype = 2
  ) +
  geom_segment(
    x = round(mean(data$conf.lcl.rescaled), 2),
    y = 0,
    yend = 0.35,
    color = "darkblue",
    lwd = 1.4,
    linetype = 2
  ) +
  annotate(
    "text",
    x = round(mean(data$conf.az.rescaled), 2),
    y = 0.40,
    label = paste("AZ =", round(mean(
      data$conf.az.rescaled
    ), 2), "(", round(sd(
      data$conf.az.rescaled
    ), 2), ")"),
    color = "firebrick"
  ) +
  annotate(
    "text",
    x = round(mean(data$conf.lcl.rescaled), 2),
    y = 0.43,
    label = paste("Local area =", round(mean(
      data$conf.lcl.rescaled
    ), 2), "(", round(sd(
      data$conf.lcl.rescaled
    ), 2), ")"),
    color = "darkblue"
  ) +
  labs(
    x = str_wrap("Confidence", width = 85),
    y = "Density",
    caption = str_wrap(
     "Mean (std.dev). Mean difference = 0.33, CI [0.27, 0.38], t = 11.613, data = 1286, p < 0.01",
     width = 95)) +
  theme_bw(base_family = "TeX Gyre Pagella")

Overlapping Distribution of Confidence in Elections for AZ and Local Area

Coefficient Plot of Treatment on Confidence by Location

In [54]:

#|   - "Confidence by Treatment Condition"
#|   - "Trust and Distrust by Treatment Condition"

# coef plot of treatment effect on confidence distinguished by location
conf.coef1 <- jtools::plot_summs(
  lm(conf.az.rescaled ~ group, data = data), 
  lm(conf.lcl.rescaled ~ group, data = data), 
  center = T, scale = T, transform.response = T,
  model.names = c("Conf.AZ", "Conf.Local"),
  coefs = c("Treatment" = "groupTreatment"), 
  colors = c("red", "black"),
  point.shape = c(8, 1))+
  ggplot2::expand_limits(x = c(-0.25, 0.25))+
  labs(y = "")+
  theme_bw(base_family = "TeX Gyre Pagella")


# Coef plots of trust and distrust models for AZ and Local items
conf.coef2 <- jtools::plot_summs(
  lm(trust.scores.az ~ group, data = data),
  lm(distrust.scores.az ~ group, data = data),
  lm(trust.scores.lcl ~ group, data = data),
  lm(distrust.scores.lcl ~ group, data = data),
  center = T, scale = T, transform.response = T,
  model.names = c("Trust.AZ", "Distrust.AZ", "Trust.Local", "Distrust.Local"),
  coefs = c("Treatment" = "groupTreatment"),
  colors = c("red", "red", "black", "black"),
  point.shape = c(8, 8, 1, 1)
)+
  ggplot2::expand_limits(x = c(-0.25, 0.25))+
  labs(y = "")+
  theme_bw(base_family = "TeX Gyre Pagella")


conf.coef1
conf.coef2

Confidence in Elections by Experiment Condition

Polychoric Item Correlation Plot

In [55]:


# Get polychoric item correlation matrix.
item.cor <- data |>
  select(q19:q24, -q23, q28_1:q28_5, q30:q35, -q34, q40_1:q40_5) |> 
  # ensure factors are ordered factors
  mutate(across(where(is.factor), ~factor(., ordered = T))) |> 
  # adjust variable column names of items
  dplyr::rename(
    trustvote_az     = q19,
    trustadmin_az    = q20,
    truststaff_az    = q21,
    trustprcss_az    = q22,
    trusttech_az     = q24,
    votefraud_az     = q28_1,
    miscount_az      = q28_2,
    turnedaway_az    = q28_3,
    foreigntamp_az   = q28_4,
    discourage_az    = q28_5,
    trustvote_lcl    = q30,
    trustadmin_lcl   = q31,
    truststaff_lcl   = q32,
    trustprcss_lcl   = q33,
    trusttech_lcl    = q35,
    votefraud_lcl    = q40_1,
    miscount_lcl     = q40_2,
    turnedaway_lcl   = q40_3,
    foreigntamp_lcl  = q40_4,
    discourage_lcl   = q40_5
  ) |> 
  correlation::correlation(
    select = 
      c("trustvote_az", "trustadmin_az", "truststaff_az", "trustprcss_az", "trusttech_az",
        "votefraud_az", "miscount_az", "turnedaway_az", "foreigntamp_az", "discourage_az"),
    select2 =
      c("trustvote_lcl", "trustadmin_lcl", "truststaff_lcl", "trustprcss_lcl","trusttech_lcl",
        "votefraud_lcl", "miscount_lcl", "turnedaway_lcl", "foreigntamp_lcl","discourage_lcl"),
    method = "polychoric") |>
  summary()

# plot item correlations
plot(item.cor) + theme(axis.text.x = element_text(angle = 45, vjust = 0.6))

Barplots

In [56]:


bar_q25 <- custom_barplot(
    data = data,
    x = q25,
    group = group
  ) +
  coord_cartesian(ylim = c(0,60), expand = T)

bar_q26 <- custom_barplot(
  data = data,
  x = q26,
  group = group
)  +
  coord_cartesian(ylim = c(0,60), expand = T)


bar_q36 <- custom_barplot(
  data = data,
  x = q36,
  group = group
)  +
  coord_cartesian(ylim = c(0,60), expand = T)

bar_q37 <- custom_barplot(
  data = data,
  x = q37,
  group = group 
) +
  coord_cartesian(ylim = c(0,60), expand = T)

#|   - "Concern for Violence in Maricopa County, AZ Elections"
#|   - "Confidence in Voter Saftey at Election Sites in Maricopa County, AZ"

bar_q25
bar_q26

#|   - "Concern for Violence in Local Area Elections"
#|   - "Confidence in Voter Saftey at Election Sites in Local Area"

bar_q36
bar_q37

In [57]:

#|   - "Concern for Violence in Maricopa County, AZ Elections"
#|   - "Confidence in Voter Saftey at Election Sites in Maricopa County, AZ"

bar_q25 <- custom_barplot(
    data = data,
    x = q25,
    group = group
  ) +
  coord_cartesian(ylim = c(0,60), expand = T)

bar_q26 <- custom_barplot(
  data = data,
  x = q26,
  group = group
)  +
  coord_cartesian(ylim = c(0,60), expand = T)


bar_q36 <- custom_barplot(
  data = data,
  x = q36,
  group = group
)  +
  coord_cartesian(ylim = c(0,60), expand = T)

bar_q37 <- custom_barplot(
  data = data,
  x = q37,
  group = group 
) +
  coord_cartesian(ylim = c(0,60), expand = T)

bars1 <- (bar_q25 + bar_q26)+
  patchwork::plot_layout(guides = "collect") &
  theme(
    text = element_text(family = "TeX Gyre Pagella"),
    legend.position = "bottom", 
    axis.title.y.left = element_blank(), 
    axis.text.y.left = element_blank(), 
    axis.ticks.y = element_blank())

bars2 <- (bar_q36 + bar_q37)+
  patchwork::plot_layout(guides = "collect") &
  theme(
    text = element_text(family = "TeX Gyre Pagella"),
    legend.position = "bottom", 
    axis.title.y.left = element_blank(), 
    axis.text.y.left = element_blank(), 
    axis.ticks.y = element_blank())

bars1
bars2

#| fig-subcap:
#|    - "Concern for Violence in Maricopa County, AZ Elections"
#|    - "Confidence in Voter Saftey at Election Sites in Maricopa County, AZ"
#|    - "Concern for Violence in Local Area Elections"
#|    - "Confidence in Voter Saftey at Election Sites in Local Area"

bar_q25 <- custom_barplot(data = data, x = q25, group = group)

bar_q26 <- custom_barplot(data = data, x = q26, group = group)

bar_q36 <- custom_barplot(data = data, x = q36, group = group)

bar_q37 <- custom_barplot(data = data, x = q37, group = group)

bar_q25
bar_q26
bar_q36
bar_q37