In xmarquez/AuthoritarianismBook: Replication Package for the Book Non-Democratic Politics by Xavier Marquez

Introduction

This vignette shows how to replicate and extend the charts in chapter 9 of my book Non-democratic Politics: Authoritarianism, Dictatorships, and Democratization (Palgrave Macmillan, 2016). It assumes that you have downloaded the replication package as follows:

if(!require(devtools)) {
  install.packages("devtools")
}

devtools::install_github('xmarquez/AuthoritarianismBook')

It also assumes you have the dplyr, ggplot2, scales, forcats, ggrepel, reshape2, and knitr packages installed:

if(!require(dplyr)) {
  install.packages("dplyr")
}

if(!require(ggplot2)) {
  install.packages("ggplot2")
}

if(!require(scales)) {
  install.packages("forcats")
}

if(!require(forcats)) {
  install.packages("scales")
}

if(!require(ggrepel)) {
  install.packages("ggrepel")
}

if(!require(reshape2)) {
  install.packages("reshape2")
}

if(!require(knitr)) {
  install.packages("knitr")
}

For the interactive graph below, you also need the plotly package:

if(!require(plotly)) {
  install.packages("plotly")
}

knitr::opts_chunk$set(fig.height = 7,
                      fig.width = 7,
                      warning = FALSE)

Figure 9.1: Growth rates by electoral regime

Original figure

This figure attempts to give a sense of the distribution of the average ranks of yearly growth rates of different regime types, 1950-2014. Since growth rates are volatile and affected by the total state of the world economy at any given time, we do not compare growth rates directly. Instead, we first calculate the growth rate percentile rank for each regime in each year (1 = highest growth in year, 0 - lowest growth in year), and then average these ranks over the entire life of the regime (so, for example, Singapore, an electoral regime with restrictions on competition, averaged a 0.7 from 1965 until 2014, a consistently high growth performer). Each violin then shows the distribution of these averages by regime type, with vertical bars indicating the bottom 10 percent, the median, and the top 10 percent of regimes (so, for example, we can see that the median democratic regime has produced growth in the top 50% of growth rates, when averaged over the life of the regime). Regime data comes from LIED [@SkaaningGerring2015], extending the PIPE dataset from @Przeworski2013; economic growth data comes from the Maddison project [@Maddison2013], the Penn World Tables [@Feenstra2013], and the World Bank. There are a total of 793 regimes in the sample (median duration 5 years, though electoral democracies in the sample tend to last longer); areas are scaled to the number of observations in each category.

library(AuthoritarianismBook)
library(dplyr)
library(ggplot2)

count_sequence_breaks <- function(seq, seq_step = 1) {
  first_diff <- c(seq_step, diff(seq)) - seq_step
  periods <- cumsum(abs(first_diff))
  periods
}

data <- economic_data %>% 
  arrange(country_name, GWn, year, variable) %>% 
  group_by(country_name, variable) %>% 
  mutate(num_years = c(NA, diff(year))) %>% 
  filter(num_years == 1, year >= 1950) %>%
  left_join(lied %>% select(-in_system)) %>%
  mutate(regime = ifelse(lexical_index %in% c(3,5,4),
                         "Electoral regime, restrictions on competition or suffrage",
                         ifelse(lexical_index %in% c(2),
                                "Unelected executive, elected legislature",
                                regime))) %>% 
  filter(!is.na(regime)) %>%
  group_by(country_name) %>%
  mutate(regime_num = count_sequence_breaks(unclass(factor(regime)), seq_step = 0)) %>%
  group_by(variable,year) %>% 
  mutate(growth_rank = percent_rank(per_capita_growth)) %>%
  ungroup() %>%
  group_by(country_name, regime_num, regime) %>%
  summarise(growth_rank = mean(growth_rank,na.rm=TRUE), 
            min_year = min(year), 
            max_year = max(year), 
            n = n(), 
            total_length = max_year - min_year)

ggplot(data = data, 
       aes(x = reorder(regime, growth_rank, FUN = "median"), 
           y = growth_rank)) + 
  geom_violin(draw_quantiles = c(0.1,0.5,0.9),
              fill="lightgrey", 
              scale="count") +
  theme_bw() +
  labs(y = "Distribution of the average rank of
       the regime's yearly growth rate \nover its entire life
       (percentile scale, \n1 means best growth performance)",
       x = "") +
  theme(legend.position = "bottom") + 
  coord_flip() +
  geom_hline(yintercept = 0.5, color = "red")

Extension: Disaggregating the data per country

There are a couple of ways of disaggregating this data. We can show the actual growth ranks:

data <- economic_data %>% 
  arrange(country_name, GWn, year, variable) %>% 
  group_by(country_name, variable) %>% 
  mutate(num_years = c(NA, diff(year))) %>% 
  filter(num_years == 1, year >= 1950) %>%
  left_join(lied %>% select(-in_system)) %>%
  mutate(regime = ifelse(lexical_index %in% c(3,5,4),
                         "Electoral regime, restrictions on competition or suffrage",
                         ifelse(lexical_index %in% c(2),
                                "Unelected executive, elected legislature",
                                regime))) %>% 
  filter(!is.na(regime)) %>%
  group_by(country_name) %>%
  mutate(regime_num = count_sequence_breaks(unclass(factor(regime)), seq_step = 0)) %>%
  group_by(variable,year) %>% 
  mutate(growth_rank = percent_rank(per_capita_growth)) %>%
  ungroup() %>%
  group_by(country_name, regime_num, regime) %>%
  mutate(min_year = min(year),
         max_year = max(year),
         n = n(),
         total_length = max_year - min_year) %>%
  ungroup()

ggplot(data = data, 
       aes(x = reorder(country_name, growth_rank),
           y = growth_rank,
           color = regime)) +
  geom_boxplot(aes(alpha = n,
                   group = paste(country_name, regime, regime_num)),
               varwidth = TRUE) +
  geom_violin(alpha = 0.3) +
  labs(x = "",
       color = "Regime type") +
  theme_bw() +
  theme(legend.position = "top") +
  guides(color = guide_legend(title.position = "top", ncol = 1)) +
  coord_flip()

Or we can plot the actual income trajectories:

data <- economic_data %>% 
  arrange(country_name, GWn, year, variable) %>% 
  filter(!is.na(per_capita), year >= 1950) %>%
  inner_join(lied %>% select(-cown:-in_system)) %>%
  mutate(regime = ifelse(lexical_index %in% c(3,5,4),
                         "Electoral regime, restrictions on competition or suffrage",
                         ifelse(lexical_index %in% c(2),
                                "Unelected executive, elected legislature",
                                regime))) %>% 
  filter(!is.na(regime)) %>%
  group_by(country_name) %>%
  mutate(regime_num = count_sequence_breaks(unclass(factor(regime)), seq_step = 0)) %>%
  group_by(variable, country_name, regime_num) %>%
  arrange(year) %>%
  mutate(time = year - first(year),
         normalized_income = per_capita/first(per_capita)) %>%
  ungroup() 

ggplot(data = data %>% filter(primary_source == "PWT8.1"),
       aes(x = time, 
           y = normalized_income,
           color = regime,
           group = paste(country_name, regime_num))) +
  geom_path() +
  geom_text(data = data %>%
              group_by(country_name, variable) %>%
              filter(time == max(time, na.rm=TRUE),
                     primary_source == "PWT8.1"),
            aes(x = time,
                y = normalized_income,
                label = paste(country_name, year)),
            check_overlap = TRUE,
            color = "black") +
  theme_bw() +
  labs(y = "Ratio of increase in GDP per capita 
       over the life of the regime",
       x = "Regime duration",
       color = "") +
  theme(legend.position = "bottom") +
  guides(color = guide_legend(title.position = "top", ncol = 1)) +
  scale_color_brewer(type = "qual", palette = "Set1") +
  scale_y_log10(labels = scales::percent) +
  facet_wrap(~variable, ncol = 1)

Figure 9.2: Distribution of the average ranks of yearly growth rates of different regime types, 1950-2010

Original figure

As in the previous figure, we first calculate the growth rate percentile rank for each regime in each year (1 = highest growth in year, 0 - lowest growth in year), and then average these ranks over the entire life of the regime. Each violin shows the distribution of these averages by regime type, with vertical bars indicating the bottom 10 percent, the median, and the top 10 percent of regimes. Regime data comes from @GeddesWrightFrantz2014; economic growth data comes from the Maddison project [@Maddison2013], the Penn World Tables [@Feenstra2013], and the World Bank. There are a total of 461 regimes in the sample (median duration 10 years, though some last much longer); areas are scaled to the number of observations in each category.

data <- economic_data %>% 
  arrange(country_name, GWn, year, variable) %>% 
  group_by(country_name, variable) %>% 
  mutate(num_years = c(NA, diff(year))) %>% 
  filter(num_years == 1,year >= 1950) %>%
  left_join(all_gwf_extended_yearly %>% 
              select(-in_system)) %>% 
  filter(!is.na(gwf_full_regimetype)) %>%
  group_by(variable, year) %>% 
  mutate(growth_rank = percent_rank(per_capita_growth)) %>%
  ungroup() %>%
  mutate(regime = ifelse(grepl("monarchy",gwf_full_regimetype),
                         "Monarchy",
                         ifelse(grepl("democracy", gwf_full_regimetype),
                                "Democracy",
                                ifelse(grepl("personal", gwf_full_regimetype),
                                       "Personalized", "Other non-democracy")))) %>%
  group_by(gwf_casename,country_name,regime) %>%
  summarise(growth_rank = mean(growth_rank, na.rm=TRUE), 
            min_year = min(year), 
            max_year = max(year), 
            n = n(), 
            total_length = max_year - min_year)

ggplot(data = data, 
       aes(x = reorder(regime, growth_rank, FUN="median"), 
           y=growth_rank)) + 
  geom_violin(draw_quantiles = c(0.1,0.5,0.9), 
              fill="lightgrey", 
              scale="count") +
  theme_bw() +
  labs(y = "Distribution of the average rank of
       the regime's yearly growth rate
       over its entire life
       (percentile scale,
       1 means best growth performance)",
       x = "") +
  theme(legend.position = "bottom") + 
  coord_flip() +
  geom_hline(yintercept = 0.5, color = "red")

Figure 9.3: Income per capita in North and South Korea

Income per capita in North and South Korea, 1950-2010. Regime data from @GeddesWrightFrantz2014. Income per capita data comes from the Maddison Project [@Maddison2013], in 1993 international dollars. Numbers for North Korea are educated guesswork; figures from 1970-1993 are likely interpolated.

data <- economic_data %>% 
  filter(primary_source == "Maddison", 
         grepl("Korea",country_name), 
         year > 1950) %>% 
  left_join(all_gwf_extended_yearly %>% select(-in_system)) %>% 
  mutate(country_name = plyr::mapvalues(country_name, 
                                        from = c("Korea, People's Republic of",
                                                 "Korea, Republic of"), 
                                        to = c("North Korea", 
                                               "South Korea"))) %>%
  mutate(num_regime = unclass(as.factor(gwf_full_regimetype))) %>%
  group_by(country_name) %>%
  mutate(period = count_sequence_breaks(num_regime, seq_step=0)) %>%
  group_by(country_name, period) %>%
  mutate(first_year = min(year))

ggplot(data = data, aes(x = year, 
                        y = per_capita, 
                        linetype = country_name,
                        color = country_name)) +
  geom_path() +
  scale_y_log10(labels = scales::dollar) +
  theme_bw() +
  labs(y = "Per capita income, 1993 international dollars
       (estimated, from Maddison)", 
       linetype = "",
       color = "") +
  theme(legend.position = "bottom") +
  geom_text(data = data %>% 
              filter(country_name == "South Korea") %>% 
              distinct(first_year, .keep_all = TRUE),
            aes(x=first_year, 
                y = per_capita, 
                label = paste(country_name, gwf_full_regimetype,sep=": ")),
            angle = 90, 
            hjust=0) +
  coord_cartesian(ylim = c(800,50000))

Figure 9.4: Repression and democracy

Repression and democracy, 1949-2013. The indicator of repression is from @Fariss2014; higher values indicate less repression. Democracy data from @Pemstein2010, extended by me [@Marquez2016].

data <- left_join(fariss %>% select(-cown:-in_system),
                  extended_uds) %>%
  mutate(country_selected = ifelse(country_name %in% 
                               c("Chile"),
                               country_name,
                               "Other"))

ggplot(data = data, 
       aes(x = adj.z1, 
           y = latentmean)) + 
  geom_point(aes(group = country_name,
                 color = forcats::fct_rev(country_selected),
                 alpha = forcats::fct_rev(country_selected))) + 
  geom_path(aes(group = country_name,
                color = forcats::fct_rev(country_selected),
                alpha = forcats::fct_rev(country_selected))) + 
  geom_smooth(method = "lm", 
              aes(group = NULL),
              color="black") + 
  geom_text(data = data %>% 
              filter(country_selected != "Other",
                     (year %in% c(min(year), 
                                 max(year))) |
                       (year %% 10 == 0)),
            aes(label = year), 
            color = "black",
            check_overlap = TRUE) + 
  geom_vline(xintercept = 0, color = "red") +
  geom_hline(yintercept = 0, color = "red") +
  theme_bw() + 
  labs(y = "Fariss latent index of repression
       (Higher is less repressive)", 
       x ="Mean UD score
       (Higher is more democratic)", 
       color = "Highlighted country",
       alpha = "Highlighted country") + 
  guides(color = guide_legend(title.position ="top"),
         alpha = guide_legend(title.position ="top")) + 
  theme(legend.position = "bottom") 

Figure 9.5: Repression in different political regimes

Repression and political regimes, 1949-2013. Shapes indicate the density of the distribution of regimes (darker means more concentration of regimes in this 2-dimensional space).The indicator of repression is from @Fariss2014; higher values indicate less repression. Regime data comes from @Kailitz2013. The printed figure in the book uses the package ggalt, which is no longer compatible with ggplot2; using ggplot2::geom_density_2d does not change anything.

library(reshape2)

data <- left_join(fariss %>% 
                    select(-cown:-in_system), 
                  extended_uds) %>%
  left_join(kailitz_yearly %>% 
              select(-cown:-in_system, -kailitz_tri) %>% 
              mutate(democracy = (grepl("democracy",
                                        combined_regime,
                                        ignore.case = TRUE)))) %>%
  melt(measure.vars = c("personal",
                        "communist",
                        "military",
                        "party",
                        "monarchy",
                        "electoral",
                        "failure",
                        "transition",
                        "democracy"), 
       na.rm = TRUE) %>%
  filter(value) %>%
  arrange(country_name, year) %>%
  rename(regime = variable) 

ggplot(data = data,
       aes(x = adj.z1, y = latentmean, group = regime)) + 
  geom_point(alpha = 0.1) +
  geom_density_2d() +
  theme_bw() +
  labs(y = "Fariss latent index of repression
       (Higher is less repressive)", 
       x = "Mean UD score
       (Higher is more democratic)", 
       alpha = "Density") + 
  theme(legend.position = "bottom") + 
  geom_vline(xintercept = 0, color = "red") +
  geom_hline(yintercept = 0, color = "red") +
  facet_wrap(~regime)

Figure 9.6: War and democracy, 1816-2003

Original figure

War and democracy, 1816-2003. War data from Gledtisch's updated list of interstate wars [@Gleditsch2004]. A war in this dataset is an armed conflict with more than 1,000 deaths. Democracy data from @Pemstein2010, extended by the author [@Marquez2016]. The printed figure in the book used the package ggalt, which is no longer compatible with ggplot2, and accounted for measurment uncertainties; using ggplot2::geom_density_2d and ignoring the uncertainties does not change anything substantal in the figure.

data <- left_join(interstate_dyads_yearly, 
                  extended_uds %>% 
                    rename(country_name.x = country_name, 
                           GWn.x = GWn, 
                           index.x = index,
                           index.pct025.x = index.pct025,
                           index.pct975.x = index.pct975) %>% 
                    select(country_name.x, 
                           GWn.x,
                           year,
                           index.x,
                           index.pct975.x,
                           index.pct025.x)) %>%
  left_join(extended_uds %>% 
                    rename(country_name.y = country_name, 
                           GWn.y = GWn, 
                           index.y = index,
                           index.pct025.y = index.pct025,
                           index.pct975.y = index.pct975) %>% 
                    select(country_name.y, 
                           GWn.y,
                           year,
                           index.y,
                           index.pct975.y,
                           index.pct025.y)) 

ggplot(data = data,
              aes(x=index.x, 
                  y=index.y)) +
  geom_point(data = data %>% 
               distinct(warname, index.x, index.y), 
             aes(x=index.x, 
                 y = index.y)) + 
  geom_density_2d() +
  theme_bw() +
  theme(legend.position = "bottom") +
  labs(x = "UD score of side A in war
       (0-1 probability scale,
       democracies = score >0.5)",
       y = "UD score of side B in war
       (0-1 probability scale
       democracies = score >0.5)", 
       alpha = "Density") +
  geom_vline(xintercept = 0.5, color = "red") +
  geom_hline(yintercept = 0.5, color = "red") +
  annotate("text", 
           x = 0.75,
           y = 0.85,
           label = "Wars between democracies
           in this quadrant
           (Most of them conflicts within WWII,
           against Finland 1941-1944)") +
  coord_cartesian(ylim = c(0,1))

Interactive Version

This requires the plotly package:

library(plotly)

data <- data %>%
  mutate(uds_side_A = index.x,
         uds_side_B = index.y,
         name = paste0(warname, 
                       " (", 
                       year,
                       ") ",
                       country_name.x,
                       " (A) vs. ",
                       country_name.y,
                       " (B)")) %>%
  select(name, year, uds_side_A, uds_side_B) %>%
  filter(!is.na(uds_side_A), !is.na(uds_side_B))

p <- ggplot(data = data,
              aes(text = name,
                  x = uds_side_A, 
                  y = uds_side_B)) +
  geom_point() + 
  theme_bw() +
  theme(legend.position="bottom") +
  labs(x = "UD score of side A in war
       (0-1 probability scale,
       democracies = score >0.5)",
       y = "UD score of side B in war
       (0-1 probability scale
       democracies = score >0.5)", 
       alpha = "Density") +
  geom_vline(xintercept = 0.5, color = "red") +
  geom_hline(yintercept = 0.5, color = "red")

ggplotly(p)

Extension 1: Wars per country

It is possible to disaggregate this data per country. The figure below puts countries that have been involved in most wars at the top:

data <- full_join(extended_uds, all_war_participants_yearly) %>%
  mutate(wartime = !is.na(type)) %>%
  group_by(country_name) %>%
  mutate(num_wars = length(unique(warname[ !is.na(warname) ]))) %>%
  ungroup() %>%
  filter(!is.na(index))

ggplot(data = data, 
       aes(x = reorder(country_name, num_wars, mean, na.rm = TRUE),
           y = year,
           alpha = wartime)) +
  geom_tile(aes(fill = index)) +
  geom_point(aes(shape = type)) +
  theme_bw() +
  theme(legend.position="top") +
  scale_fill_gradient2(midpoint = 0.5) +
  labs(x = "",
       fill = "Unified democracy score",
       alpha = "Wartime?") +
  coord_flip()

References

xmarquez/AuthoritarianismBook documentation built on May 4, 2019, 1:24 p.m.