.Rhistory

"3" = "somewhat interest in politics",
"4" = "very interest in politics")))
ggplot(reg_df, aes(x=sev, y=pol_trust)) +
geom_jitter(width=0.1, alpha=0.2) +
geom_smooth() +
facet_grid(. ~ pol_interest,
labeller=labeller(pol_interest = c("1" = "not at all interested in politics",
"2" = "not very interest in politics",
"3" = "somewhat interest in politics",
"4" = "very interest in politics")))
ggplot(reg_df, aes(x=sev, y=pol_trust)) +
geom_jitter(width=0.1, alpha=0.2) +
geom_smooth() +
facet_grid(. ~ pol_interest,
labeller=labeller(pol_interest = c("1" = "not at all interested in politics",
"2" = "not very interest in politics",
"3" = "somewhat interest in politics",
"4" = "very interest in politics"))) +
theme(plot.subtitle = element_text(size = 13),
plot.title = element_text(size = 15, face = "bold")) +
labs(title = "Political trust in national government vs. self-expression values, split by political interest",
x = "self-expression values index score",
y = "political trust", subtitle = "data: Asian Barometer Survey, wave 4, 2015/6")
ggplot(reg_df, aes(x=sev, y=pol_trust)) +
geom_jitter(width=0.1, alpha=0.2) +
geom_smooth() +
facet_grid(. ~ pol_interest,
labeller=labeller(pol_interest = c("1" = "not at all interested in politics",
"2" = "not very interested in politics",
"3" = "somewhat interested in politics",
"4" = "very interested in politics"))) +
theme(plot.subtitle = element_text(size = 13),
plot.title = element_text(size = 15, face = "bold")) +
labs(title = "Political trust in national government vs. self-expression values, split by political interest",
x = "self-expression values index score",
y = "political trust", subtitle = "data: Asian Barometer Survey, wave 4, 2015/6")
ggplot(reg_df, aes(x=sev, y=pol_trust)) +
geom_jitter(width=0.1, alpha=0.2) +
geom_smooth() +
facet_grid(. ~ pol_interest,
labeller=labeller(pol_interest = c("1" = "not at all interested in politics",
"2" = "not very interested in politics",
"3" = "somewhat interested in politics",
"4" = "very interested in politics"))) +
theme(plot.subtitle = element_text(size = 13),
plot.title = element_text(size = 15, face = "bold")) +
labs(title = "Political trust in national government vs. self-expression values (added noise), split by political interest",
x = "self-expression values index score",
y = "political trust", subtitle = "data: Asian Barometer Survey, wave 4, 2015/6")
plot_bivariate <- ggplot(reg_df, aes(x=sev, y=pol_trust)) +
geom_jitter(width=0.1, alpha=0.2) +
geom_smooth() +
facet_grid(. ~ pol_interest,
labeller=labeller(pol_interest = c("1" = "not at all interested in politics",
"2" = "not very interested in politics",
"3" = "somewhat interested in politics",
"4" = "very interested in politics"))) +
theme(plot.subtitle = element_text(size = 13),
plot.title = element_text(size = 15, face = "bold")) +
labs(title = "Political trust in national government vs. self-expression values (added noise), split by political interest",
x = "self-expression values index score",
y = "political trust", subtitle = "data: Asian Barometer Survey, wave 4, 2015/6")
ggsave('bivariate.png', path="./plots", plot=plot_trend, width=10, height=5, device="png")
ggsave('trend.png', path="./plots", plot=plot_trend, width=7.5, height=5, device="png")
ggsave('bivariate.png', path="./plots", plot=plot_bivariate, width=10, height=5, device="png")
ggsave('bivariate.png', path="./plots", plot=plot_bivariate, width=10, height=3.5, device="png")
plot_bivariate <- ggplot(reg_df, aes(x=sev, y=pol_trust)) +
geom_jitter(width=0.1, alpha=0.2) +
geom_smooth() +
facet_grid(. ~ pol_interest,
labeller=labeller(pol_interest = c("1" = "not at all interested in politics",
"2" = "not very interested in politics",
"3" = "somewhat interested in politics",
"4" = "very interested in politics"))) +
theme(plot.subtitle = element_text(size = 13),
plot.title = element_text(size = 15, face = "bold")) +
labs(title = "Political trust in national government vs. self-expression values",
x = "self-expression values index score",
y = "political trust", subtitle = "added noise, split by political interest [data: Asian Barometer Survey, wave 4, 2015/6]")
ggsave('bivariate.png', path="./plots", plot=plot_bivariate, width=10, height=3.5, device="png")
ggsave('bivariate.png', path="./plots", plot=plot_bivariate, width=10, height=5, device="png")
ggsave('bivariate.png', path="./plots", plot=plot_bivariate, width=10, height=3.5, device="png")
######################################################
#
#    Political trust in national government in China
#    am empirical analysis using data by the Asian Barometer Survey
#
#    R script
#
#####################################################@
# Environment preparation ------------------------------------------------------------
# Clearing memory, setting language to English
rm(list = ls())
Sys.setenv(LANG = "en")
setwd(here::here())
# Loading packages
packages <- c(
"haven", #reading SPSS data
"readr", #dependency of haven
"here", #setting wd dynamically to root of current R project file. thus requires the use of RStudio projects alongside this script
"dplyr", #data manipulation
"Hmisc", #necessary for label() due to SPSS data
"ggplot2", #used for plots
"labelled", #necessary to remove labels
"naniar", #replacing specific values with NA
"data.table", #used for the setnames() function
"psych", #package necessary for the reverse.code() function
"flextable", #used to create summary table and to convert tables to docx
"gtsummary", #used to create summary tables
"stargazer" #used to create regression tables
)
installed_packages <- packages %in% rownames(installed.packages())
if (any(installed_packages == FALSE)) { # Install packages not yet installed
install.packages(packages[!installed_packages])
}
lapply(packages, library, character.only = TRUE) # Packages loading
# Loading and preparing Asian Barometer Survey data ------------------------------------------------------------
abs_wave1 <- read_sav(here("data_ABS", "ABSdata_PRC_wave1", "Mainland v4.2.sav"))
abs_wave2 <- read_sav(here("data_ABS", "ABSdata_PRC_wave2", "04_China.sav"))
abs_wave3 <- read_sav(here("data_ABS", "ABSdata_PRC_wave3", "ABS3w.CN.sav"))
abs_wave4 <- read_sav(here("data_ABS", "ABSdata_PRC_wave4", "W4_China_merged20181206.SAV"))
# View(abs_wave1)
# View(abs_wave2)
# View(abs_wave3)
# View(abs_wave4)
# Data transformation functions -------------------------------------------
#note that this function will only work on matrices, it will NOT work on data frames
rev <- function(data,variable) {
data <- psych::reverse.code(
keys = variable,
items = data)#,
#mini = min(data[,variable]), #apparently, reverse.code() finds this by default, but I will keep the code here
#maxi = max(data[,variable])) #apparently, reverse.code() finds this by default, but I will keep the code here
colnames(data)[which(colnames(data) == paste0(variable,"-"))] <- variable
return(data)
}
#supply a dataframe (data_list) and a CHARACTER vector (variable_list) that contains all the names of the vars to be reversed
rev_list <- function(data_list,variable_list,data,variable,log=TRUE) {
for (item in variable_list) {
data_list <- rev(
data=data_list,
variable=item)
if(log==TRUE){
print(paste0(
"Variable reversed: ", item, " (range is ", min(data_list[,item], na.rm = TRUE),  ":", max(data_list[,item], na.rm = TRUE), ")"
))}
next
}
return(data_list)
}
rplc_NA <- function(data_list, variable_list, values,log=TRUE) {
for (item in variable_list) {
data_list <- naniar::replace_with_na(data_list, replace = list(item = values))
if(log==TRUE){
print(paste0(
"NA inserted for: ", item, " with values ", values
))}
next
}
}
# Time-series analysis: variable preparation -------------------------------------------
#these next few lines import the political trust variable from abs wave 1, 2, 3, and 4 respectively
#abs3 and 4 need to be recoded to match the rising likert scale of the other two variables
#first variable
trust_abs1 <- data.frame(pol_trust = abs_wave1$q008) %>%
remove_labels() %>%
naniar::replace_with_na(replace = list(pol_trust = c(0,98,99)))
trust_abs1$year = "2002"
#second variable
trust_abs2 <- data.frame(pol_trust = abs_wave2$q008) %>%
remove_labels() %>%
naniar::replace_with_na(replace = list(pol_trust = c(7,8,9)))
trust_abs2$year = "2007/8"
#third variable
trust_abs3 <- data.frame(pol_trust = abs_wave3$q9,
dummy = c(rep(NA,length(abs_wave3$q9)))) %>% #adding a dummy var b/c reverse.code() somehow doesn't work with atomic vectors
remove_labels() %>%
naniar::replace_with_na(replace = list(pol_trust = c(7,8,9)))
trust_abs3 <- rev_list(data_list=trust_abs3, variable_list="pol_trust")
trust_abs3 <- as.data.frame(trust_abs3)
trust_abs3$year = "2011"
trust_abs3 <- trust_abs3[,c(1,3)] #this gets rid of the dummy variable
#fourth variable
trust_abs4 <- data.frame(pol_trust = abs_wave4$q9,
dummy = c(rep(NA,length(abs_wave4$q9)))) %>% #adding a dummy var b/c reverse.code() somehow doesn't work with atomic vectors
remove_labels() %>%
naniar::replace_with_na(replace = list(pol_trust = c(-1,7,8,9))) %>%
rev_list(variable_list="pol_trust")
trust_abs4 <- as.data.frame(trust_abs4)
trust_abs4$year = "2015/6"
trust_abs4 <- trust_abs4[,c(1,3)] #this gets rid of the dummy variable
# Time-series analysis: data frame creation -------------------------------
#The next few lines combine the survey responses to the question about trust in national government
#After that, some summary statistics are created, such as count or proportion
#Multiple data frames are created to serve different visualization purposes
#first data frame
time_series_trust_raw <- rbind(trust_abs1,
trust_abs2,
trust_abs3,
trust_abs4)
#second data frame
time_series_trust_avg <- time_series_trust_raw %>%
group_by(year) %>%
summarise(avg = sprintf("%.2f", mean(pol_trust, na.rm = TRUE))) %>%
mutate(avg = as.numeric(avg))
#third data frame
time_series_trust <- time_series_trust_raw %>%
group_by(year, pol_trust) %>%
summarise(count = n()) %>%
left_join(y=time_series_trust_avg, by="year") %>%
group_by(year) %>%
summarise(pol_trust = pol_trust,
count = count,
avg = avg,
year_total = sum(count),
year_prop = count / year_total) %>%
mutate(
dummy = "dummy",
avg = avg,
year = as.factor(year),
pol_trust = factor(recode(pol_trust,
"1" = "no trust at all",
"2" = "not very much trust",
"3" = "quite a lot of trust",
"4" = "a great deal of trust"),
ordered = TRUE,
levels = c(
NA,
"a great deal of trust",
"quite a lot of trust",
"not very much trust",
"no trust at all"),
exclude = NULL
)
)
# Time-series analysis: visualization -------------------------------------
#This creates a table that displays the number of responses
table_freq <-
unclass(
table(
mutate(time_series_trust_raw,
pol_trust = factor(recode(pol_trust,
"1" = "no trust at all",
"2" = "not very much trust",
"3" = "quite a lot of trust",
"4" = "a great deal of trust"),
ordered = TRUE,
levels = c(
NA,
"a great deal of trust",
"quite a lot of trust",
"not very much trust",
"no trust at all"),
exclude = NULL
)
),
useNA="ifany")
)
vars <- "missing"
vars[2:5] <- row.names(table_freq)[2:5]
vars
table_freq <- as.data.frame(table_freq)
table_freq$Response <- vars
table_freq <- table_freq[,c("Response",
names(table_freq)[names(table_freq) != "Response"])]
rownames(table_freq) <-  NULL
table_freq <- flextable::flextable(table_freq)
#This creates a bar plot that visualizes the survey responses across all years
plot_prop <- ggplot(data=time_series_trust) +
geom_bar(mapping=aes(x=year, y=year_prop, fill=pol_trust), stat = "identity", color = "black") +
scale_fill_manual(aesthetics="fill",
values=c(rgb(0.1,0.1,1.0,0.6),
rgb(0.3,0.0,0.7,0.6),
rgb(0.7,0.0,0.4,0.6),
rgb(1.0,0.0,0.2,0.6))) +
theme(plot.subtitle = element_text(size = 12),
panel.grid.minor = element_line(linetype = "blank"),
plot.title = element_text(size = 15)) +labs(title = "Political trust in national government in China",
x = NULL, y = "proportion of total survey responses",
fill = NULL, subtitle = "survey response breakdown by year [data: Asian Barometer Survey, waves 1-4]")
#This creates a bar plot that visualizes the averages across all years
plot_trend <- ggplot(data=time_series_trust_avg) +
geom_bar(mapping=aes(x=year, y=avg), stat="identity", fill=rgb(0.6,0.6,0.6,0.6)) +
geom_text(mapping=aes(x=year, y=avg, label=avg),  position=position_dodge(width=0.9), vjust=2.5, size=6) +
coord_cartesian(ylim=c(1,4)) +
theme(plot.subtitle = element_text(size = 11),
panel.grid.minor = element_line(linetype = "blank"),
plot.title = element_text(size = 15)) +labs(title = "Political trust in national government in China",
x = NULL, y = "average", subtitle = "average survey responses per year [data: Asian Barometer Survey, waves 1-4]")
# Regression analysis: data frame creation --------------------------------------------------------------
reg_df <- data.frame(abs_wave4) %>%
select(se3_2, se2, se5a, se9, ir13, q1, q41, q118, q155, q161, q62, q57, q49, q148, q76, q81, q24, se7a, q44) %>%
cbind(pol_trust = trust_abs4$pol_trust) %>%
remove_labels() %>%
naniar::replace_with_na_at(.vars = c("se3_2", "se2","ir13"),
condition = ~.x %in% -1) %>%
naniar::replace_with_na_at(.vars = "se5a",
condition = ~.x %in% c(-1, 90, 98, 99)) %>%
naniar::replace_with_na_at(.vars = "se9",
condition = ~.x %in% c(-1, 9)) %>%
naniar::replace_with_na_at(.vars = "se7a",
condition = ~.x %in% c(-1, 0, 8, 9)) %>%
naniar::replace_with_na_at(.vars = c("q118", "q44"),
condition = ~.x %in% c(-1, 8, 9)) %>%
naniar::replace_with_na_at(.vars = c("q1", "q155", "q161", "q62", "q57", "q148", "q76", "q81", "q24", "q9"),
condition = ~.x %in% c(-1, 7, 8, 9)) %>%
naniar::replace_with_na_at(.vars = "q41",
condition = ~.x %in% c(-1, 5, 7, 8, 9)) %>%
naniar::replace_with_na_at(.vars = "q49",
condition = ~.x %in% c(-1, 97, 98, 99)) %>%
mutate(ir13 = recode(ir13, "1" = "1", "2" = "1", "3" = "1", "4" = "2"),
q76 = recode(q76, "1" = "1", "2" = "2.666", "3" = "4"),
q81 = recode(q81, "1" = "1", "2" = "4"))
reg_df <- as.data.frame(sapply(reg_df, as.numeric))
reg_df <- rev_list(data_list = reg_df,
variable_list = c("q81",
"q76",
"q161",
"q62",
"q57",
"q49",
"q148",
"q24",
"q44",
"q1"))
reg_df <- as.data.frame(reg_df)
setnames(reg_df,
old =c(
"se3_2", "se2", "se5a", "se9","ir13", #demographic controls
"q1", "q41", "q118", "q155", #regime performance
"q161", "q62", "q57", #cultural factors
"q49", #internet use
"q148", "q76", "q81", "q24", "se7a", #self-expression values components
"q44" #political interest
),
new =c(
"age", "female", "education_y", "unemployed","rural", #demographic controls
"economic_perf", "healthcare", "corruption", "unfairness_income_dist", #regime performance
"patriotism", "respect_for_authority", "allocentric_selfinterest", #cultural factors
"internet_use", #internet use
"diversity_tolerance", "civic_protest", "free_media", "general_trust", "weak_religiousness", #self-expression values components
"pol_interest" #political interest
)
)
reg_df$sev <- as.numeric(sprintf("%.2f",
rowMeans(reg_df[,c("diversity_tolerance", "civic_protest", "free_media", "general_trust", "weak_religiousness")], na.rm=TRUE)))
reg_df <- mutate(reg_df,
moderation_effect = pol_interest*sev)
# Regression analysis: model calculation ---------------------------------------------
#the following lines set up name vectors to make the regression syntax tidier
demographic_controls <- c("age", "female", "education_y", "unemployed","rural")
regime_performance <- c("economic_perf", "healthcare", "corruption", "unfairness_income_dist")
cultural_factors <- c("patriotism", "respect_for_authority", "allocentric_selfinterest")
internet_use <- "internet_use"
sev <- "sev"
sev_components <- c("diversity_tolerance", "civic_protest", "free_media", "general_trust", "weak_religiousness")
pol_interest <- "pol_interest"
moderation_effect <- "moderation_effect"
pol_trust <- "pol_trust"
m1 <- lm(
reformulate(
termlabels = c(
demographic_controls,
regime_performance,
cultural_factors,
internet_use,
sev),
response = pol_trust),
data = reg_df)
m2 <- lm(
reformulate(
termlabels = c(
demographic_controls,
regime_performance,
cultural_factors,
internet_use,
sev,
pol_interest),
response = pol_trust),
data = reg_df)
m3 <- lm(
reformulate(
termlabels = c(
demographic_controls,
regime_performance,
cultural_factors,
internet_use,
moderation_effect),
response = pol_trust),
data = reg_df)
m4 <- lm(
reformulate(
termlabels = c(
demographic_controls,
regime_performance,
cultural_factors,
internet_use,
sev,
pol_interest,
moderation_effect),
response = pol_trust),
data = reg_df)
m5 <- lm(
reformulate(
termlabels = c(
demographic_controls,
regime_performance,
cultural_factors,
internet_use,
sev_components),
response = pol_trust),
data = reg_df)
m6 <- lm(
reformulate(
termlabels = c(
demographic_controls,
regime_performance,
cultural_factors,
internet_use,
sev_components,
pol_interest),
response = pol_trust),
data = reg_df)
# Regression analysis: visualizations -------------------------------------
#this creates a table that summarizes the original, untransformed variables from ABS wave 4
table_summary0 <- gtsummary::tbl_summary(haven::as_factor(abs_wave4[,c("se3_2", "se2", "se5a", "se9", "ir13",
"q1", "q41", "q118", "q155", "q161",
"q62", "q57", "q49", "q148", "q76",
"q81", "q24", "se7a", "q44", "q9")])) %>%
bold_labels()
#this creates a table that summarizes the variables in the regression data frame
table_summary1 <- gtsummary::tbl_summary(reg_df) %>%
bold_labels()
dir.create("./plots")
#this creates a more concise table that summarizes the variables in the regression data frame
tabe_summary2 <- stargazer::stargazer(reg_df, type="html", out="plots/summary.html", omit.summary.stat = c("p25","p75"), digits=2,
covariate.labels = c(
"Age", "Female", "Formal education years", "Is unemployed","Rural residence",
"Economic performance", "Healthcare", "Corruption", "Unfairness of income distribution",
"Patriotism", "Respect for authority", "Allocentric self-interest",
"Internet use",
"(SEV) Tolerance for diversity", "(SEV) Inclination to civic protest", "(SEV) Preference for free media", "(SEV) Generalized social trust","(SEV) Weak religiousness",
"Political interest",
"Political trust",
"Self-expression values index",
"SE values * political interest"
))
tabe_reg1 <- stargazer::stargazer(m3,m4,m1,m2,
type="html",
title="OLS model, dependent variable: political trust in national government",
out="plots/regression1.html",
dep.var.caption = "",
dep.var.labels.include=FALSE,
single.row = TRUE,
covariate.labels = c(
"Age", "Female", "Formal education years", "Is unemployed","Rural residence",
"Economic performance", "Healthcare", "Corruption", "Unfairness of income distribution",
"Patriotism", "Respect for authority", "Allocentric self-interest",
"Internet use",
"Self-expression values index",
"Political interest",
"SE values * political interest"
))
plot_bivariate
a <- c("1" = "not at all interested in politics",
"2" = "not very interest in politics",
"3" = "somewhat interest in politics",
"4" = "very interest in politics")
b <- c("abc","dcgd")
table_reg2 <- stargazer::stargazer(m5,m6,
type="html",
title="OLS model, dependent variable: political trust in national government",
out="plots/regression2.html",
dep.var.caption = "",
dep.var.labels.include=FALSE,
single.row = TRUE,
covariate.labels = c(
"Age", "Female", "Formal education years", "Is unemployed","Rural residence",
"Economic performance", "Healthcare", "Corruption", "Unfairness of income distribution",
"Patriotism", "Respect for authority", "Allocentric self-interest",
"Internet use",
"(SEV) Tolerance for diversity", "(SEV) Inclination to civic protest", "(SEV) Preference for free media", "(SEV) Generalized social trust","(SEV) Weak religiousness",
"Political interest"
))
plot_bivariate <- ggplot(reg_df, aes(x=sev, y=pol_trust)) +
geom_jitter(width=0.1, alpha=0.2) +
geom_smooth() +
facet_grid(. ~ pol_interest,
labeller=labeller(pol_interest = c("1" = "not at all interested in politics",
"2" = "not very interested in politics",
"3" = "somewhat interested in politics",
"4" = "very interested in politics"))) +
theme(plot.subtitle = element_text(size = 13),
plot.title = element_text(size = 15, face = "bold")) +
labs(title = "Political trust in national government vs. self-expression values",
x = "self-expression values index score",
y = "political trust", subtitle = "added noise, split by political interest [data: Asian Barometer Survey, wave 4, 2015/6]")
# Plot and table output ---------------------------------------------------
table_freq
plot_prop
plot_trend
table_summary1
plot_bivariate
save_as_docx(table_freq, path="plots/freq.docx")
ggsave('prop.png', path="./plots", plot=plot_prop, width=7.5, height=5, device="png")
ggsave('trend.png', path="./plots", plot=plot_trend, width=7.5, height=5, device="png")
ggsave('bivariate.png', path="./plots", plot=plot_bivariate, width=10, height=3.5, device="png")
table_summary0 %>% as_flex_table() %>% flextable::save_as_docx(path="plots/summary0.docx")
table_summary1 %>% as_flex_table() %>% flextable::save_as_docx(path="plots/summary1.docx")
table_summary2
table_reg1
table_reg2