ジニ指数と所得分布 - 続編

南部アフリカ諸国の貧困と不平等の現状

1月9日（火）のジニ指数の演習ファイルをもとに、1月11日（木）の講義に関係したグラフをいくつか作成し、以下に提示します。基本的に、木曜日の講義を聴きながら作成したものです。

世界開発指標（World Development Indicators）[Link]

貧困と不平等（Poverty and Inequality）

所得または消費の分配（Distribution of income or consumption）

GINI 指数 (世界銀行推計)：SI.POV.GINI [Link]

下位 10% が占める所得シェア：SI.DST.FRST.10 [Link]

下位 20% が占める所得シェア：SI.DST.FRST.20 [Link]

2番目の 20% が占める収入シェア：SI.DST.02ND.20 [Link]

3番目の 20% が占める収入シェア ：SI.DST.03RD.20 [Link]

4番目の 20% が占める収入シェア：SI.DST.04TH.20 [Link]

上位 20% が占める収入シェア：SI.DST.05TH.20 [Link]

上位 10% が占める収入シェア：SI.DST.10TH.10 [Link]

概要を把握するための作業

準備

library(tidyverse)
library(WDI)
library(DescTools)

データの読み込み

地域情報を利用するために、extra = TRUE を加えました。

df_gini_extra <- WDI(indicator = c(gini = "SI.POV.GINI",
                            `0-10` = "SI.DST.FRST.10",
                            `0-20` = "SI.DST.FRST.20",
                            `20-40` = "SI.DST.02ND.20",
                            `40-60` = "SI.DST.03RD.20",
                            `60-80` = "SI.DST.04TH.20",
                            `80-100` = "SI.DST.05TH.20",
                            `90-100` = "SI.DST.10TH.10"), extra = TRUE)

保存と読み込み

何度もダウンロードしなくて良いように、保存したものを読み込みます。

write_csv(df_gini_extra, "data/gini_extra.csv")

df_gini_extra <- read_csv("data/gini_extra.csv")

Rows: 16758 Columns: 20── Column specification ─────────────────────────────────────────────────────────────────────────
Delimiter: ","
chr   (7): country, iso2c, iso3c, region, capital, income, lending
dbl  (11): year, gini, 0-10, 0-20, 20-40, 40-60, 60-80, 80-100, 90-100, longitude, latitude
lgl   (1): status
date  (1): lastupdated
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

データの確認

df_gini_extra

必要な変数のみ抽出

df_gini_extra <- df_gini_extra |> select(country, iso2c, year, gini:region)

gini が 欠損値（NA） ではない値のもののみ表示

df_gini_extra |> drop_na(gini)

gini の値の大きい順（降順）に並び替え

df_gini_extra |> drop_na(gini) |> 
  arrange(desc(gini)) |> distinct(country, year, gini, region)

それぞれの国の最も新しいデータだけを取り降順に並べます

df_gini_extra_recent <-df_gini_extra |> drop_na(gini) |> group_by(country) |> filter(year == max(year)) |>
  arrange(desc(gini))
df_gini_extra_recent

変形：各階級の値を縦に並べ、縦長形式の表にします。Long format

pivot_longer(cols, names_to = "", values_to = "") についてはいずれ説明します。ここでは、レベルに分けられたものを levels という名の列にレベルを、value という名の列に、その値を並べたものとします。

df_gini_extra_long <- df_gini_extra |> 
  pivot_longer(`0-10`:`90-100`, names_to = "levels", values_to = "value")
df_gini_extra_long

日本の情報の確認

df_gini_extra_long |> filter(country == "Japan") |> 
  drop_na(gini) |> distinct(country, year, gini, levels, value)

COUNTRIES_D <- c("Japan","United States", "South Africa")

国ごとの最近の GINI 指数（降順）

df_gini_extra |> select(country, year, gini:`90-100`) |> filter(country %in% COUNTRIES_D) |> 
  drop_na(gini) |> group_by(country) |> filter(year == max(year))

何種類かの棒グラフで表示

df_gini_extra_long |> filter(country %in% COUNTRIES_D) |>
  drop_na(gini) |> group_by(country) |> filter(year == max(year)) |> 
  ggplot(aes(factor(country, levels = COUNTRIES_D), value, fill = levels)) + geom_col(position = "dodge", col = "black", linewidth = 0.1) +
   geom_text(aes(group = levels, label = value), vjust = -0.2, position = position_dodge(width = 0.9)) + 
  labs(title = "Distribution of Wealth in Three Countries in Recent Year", x = "")

df_gini_extra_long |> filter(country %in% COUNTRIES_D) |>
  drop_na(gini) |> group_by(country) |> filter(year == max(year)) |>
  filter(!(levels %in% c('0-10','90-100'))) |> 
  ggplot(aes(factor(country, levels = COUNTRIES_D), value, fill = levels)) + geom_col(position = "dodge", col = "black", linewidth = 0.1) +
   geom_text(aes(group = levels, label = value), vjust = -0.2, position = position_dodge(width = 0.9)) + 
  labs(title = "Distribution of Wealth in Three Countries in Recent Year", x = "")

考察：0-10, 90-100 はなくても良いかもしれない。

df_gini_extra_long |> filter(country %in% COUNTRIES_D) |> filter(year == 2010) |>
  drop_na(gini) |> ggplot(aes(factor(country, levels = COUNTRIES_D), value, fill = levels)) + geom_col(position = "dodge", col = "black", linewidth = 0.1) +
   geom_text(aes(group = levels, label = value), vjust = -0.2, position = position_dodge(width = 0.9)) + 
  labs(title = "Distribution of Wealth in Three Countries in 2010", x = "")

考察：データが限られているので、年を揃えるのは難しい。最新のデータのみを使う

df_gini_extra_long |> filter(country %in% COUNTRIES_D) |>
  drop_na(gini) |> group_by(country) |> filter(year == max(year)) |>
  filter(!(levels %in% c('0-10','90-100'))) |> 
  ggplot(aes(factor(country, levels = COUNTRIES_D), value, fill = levels)) + geom_col(position = "dodge", col = "black", linewidth = 0.1) +
   geom_text(aes(group = levels, label = value), vjust = -0.2, position = position_dodge(width = 0.9)) + 
  labs(title = "Distribution of Wealth in Three Countries in Recent Year", x = "")

ローレンツ曲線を表示する準備

Derivation of the Lorenz curve and Gini coefficient for global income in 2011 [リンク]

df_gini_calc_recent <- df_gini_extra_recent |> 
  mutate(`0` = 0, `10` = `0-10`, `20` = `0-20`,
         `30` = `0-20`+`20-40`/2, `40` = `0-20` + `20-40`, 
         `50` = `0-20` + `20-40` + `40-60`/2, 
         `60` = `0-20` + `20-40` + `40-60`, 
         `70` = `0-20` + `20-40` + `40-60` + `60-80`/2,  
         `80` = `0-20` + `20-40` + `40-60` + `60-80`, 
         `90` = `0-20` + `20-40` + `40-60` + `60-80` + `80-100`-`90-100`,
         `100` = 100) |>
  select(-c(`0-10`:`90-100`)) # 不必要な部分を消去
df_gini_calc %>% drop_na() 

縦長の Long Table に変換

df_gini_calc_recent_long <- df_gini_calc_recent |>  pivot_longer(`0`:`100`, names_to = "classes", values_to = "cumulative_share") |> mutate(classes = as.numeric(classes))
df_gini_calc_long %>% drop_na() 

確認

df_gini_calc_recent_long |> filter(country == "Japan") |> 
  ggplot() + 
  geom_line(aes(classes, cumulative_share)) + 
  geom_segment(aes(x = 0, y = 0, xend = 100, yend = 100), color = 'red') + 
  scale_x_continuous(breaks = seq(0,100,by=20)) + 
  scale_y_continuous(breaks = seq(0,100,by=20)) #+
  #annotate("text", x = 10, y = 80, label = gini)

世界の状況

それぞれの国の最新のデータ

ジニ指数の降順

df_gini_extra_recent <-df_gini_extra |> drop_na(gini) |> group_by(country) |> filter(year == max(year)) |>
  arrange(desc(gini))
df_gini_extra_recent

ジニ指数の大きな30カ国を棒グラフにします

top30gini <- df_gini_extra_recent |>  
  arrange(desc(gini)) |> head(30) |> pull(country)
df_gini_extra_recent |> filter(country %in% top30gini) |> 
  ggplot(aes(factor(country, levels = top30gini), gini, fill = region)) + geom_col() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 1, hjust=1), legend.position = "top") +
  labs(title = "Top 30 Countries of Recent Gini Index", x = "")

考察：二地域に限られています

分布を地域ごとの箱ひげ図で見てみます

箱ひげ図とは？ [リンク]

df_gini_extra_recent |> drop_na(region) |> filter(region != "Aggregates") |>
  ggplot(aes(gini, region, fill = region)) + geom_boxplot() +
  theme(legend.position = "")

南アフリカ５カ国について

SOUTH_AFRICA_FIVE <- c("South Africa", "Namibia", "Eswatini", "Botswana", "Lesotho")

経年変化を表す折れ線グラフ

df_gini_extra |> filter(country %in% SOUTH_AFRICA_FIVE ) |>
  drop_na(gini) |>
  ggplot(aes(year, gini, col = factor(country, level = SOUTH_AFRICA_FIVE))) + 
  geom_line() + labs(title = "Gini Index of Five Countries", col = "From gini top")

df_gini_extra_long |> filter(country %in% SOUTH_AFRICA_FIVE ) |>
  drop_na(value) |>
  ggplot(aes(year, value, col = levels)) + geom_line() + facet_wrap(~factor(country, level = SOUTH_AFRICA_FIVE)) + labs(title = "Change of ratio of each level")

早くからデータがある国のジニ指数が低いように見える

最新のデータ

データの数も少ないので、最新のデータのみに限る

df_gini_extra_recent |> filter(country %in% SOUTH_AFRICA_FIVE)

SOUTH_AFRICA_FIVE の国の順番を、Gini 指数の大きい順に並べておく

ローレンツ曲線

df_gini_calc_recent_long |> filter(country %in% SOUTH_AFRICA_FIVE) |> 
  ggplot() + 
  geom_line(aes(classes, cumulative_share, col = factor(country, levels = SOUTH_AFRICA_FIVE))) + 
  geom_segment(aes(x = 0, y = 0, xend = 100, yend = 100, col = country), color = 'red') + 
  scale_x_continuous(breaks = seq(0,100,by=20)) + 
  scale_y_continuous(breaks = seq(0,100,by=20)) +
  labs(col = "From gini top")

所得の割合を表す棒グラフ

df_gini_extra_long |> filter(country %in% SOUTH_AFRICA_FIVE) |>
  drop_na(gini) |> group_by(country) |> filter(year == max(year)) |>
  filter(!(levels %in% c('0-10','90-100'))) |> 
  ggplot(aes(factor(country, levels = SOUTH_AFRICA_FIVE), value, fill = levels)) + geom_col(position = "dodge", col = "black", linewidth = 0.1) +
   geom_text(aes(group = levels, label = value), vjust = -0.2, position = position_dodge(width = 0.9)) + 
  labs(title = "Distribution of Wealth in Five Countries in Recent Year", x = "")

ジニ指数と Top 20% の値の相関

df_gini_extra_recent |> filter(region != "Aggregates") |> drop_na(`80-100`) |>
  ggplot(aes(gini, `80-100`)) + geom_point(aes(col = region)) + 
  geom_smooth(formula = 'y ~ x', method = "lm")

考察：かなり強い正の相関があり、回帰直線がかなり適合している。

練習： 国のリストを変えてみよう

国の選択

CHOSEN_GINI_COUNTRIES <- c("Suriname", "Belize", "Brazil", "Colombia")

経年変化を表す折れ線グラフ

df_gini_extra |> filter(country %in% CHOSEN_GINI_COUNTRIES) |>
  drop_na(gini) |>
  ggplot(aes(year, gini, col = factor(country, level = CHOSEN_GINI_COUNTRIES))) + 
  geom_line() + labs(title = "Gini Index of Chosen Countries", col = "From gini top")

df_gini_extra_long |> filter(country %in% CHOSEN_GINI_COUNTRIES) |>
  drop_na(value) |>
  ggplot(aes(year, value, col = levels)) + geom_line() + facet_wrap(~factor(country, level = CHOSEN_GINI_COUNTRIES)) + labs(title = "Change of ratio of each level")

考察：

最新のデータ

データの数も少ないので、最新のデータのみに限る

df_gini_extra_recent |> filter(country %in% CHOSEN_GINI_COUNTRIES)

考察：

CHOSEN_GINI_COUNTRIES の国の順番を、Gini 指数の大きい順に並べておく

ローレンツ曲線

df_gini_calc_recent_long |> filter(country %in% CHOSEN_GINI_COUNTRIES) |> 
  ggplot() + 
  geom_line(aes(classes, cumulative_share, col = factor(country, levels = CHOSEN_GINI_COUNTRIES))) + 
  geom_segment(aes(x = 0, y = 0, xend = 100, yend = 100, col = country), color = 'red') + 
  scale_x_continuous(breaks = seq(0,100,by=20)) + 
  scale_y_continuous(breaks = seq(0,100,by=20)) +
  labs(col = "From top gini")

考察：

所得の割合を表す棒グラフ

df_gini_extra_long |> filter(country %in% CHOSEN_GINI_COUNTRIES) |>
  drop_na(gini) |> group_by(country) |> filter(year == max(year)) |>
  filter(!(levels %in% c('0-10','90-100'))) |> drop_na(value) |>
  ggplot(aes(factor(country, levels = CHOSEN_GINI_COUNTRIES), value, fill = levels)) + geom_col(position = "dodge", col = "black", linewidth = 0.1) +
   geom_text(aes(group = levels, label = value), vjust = -0.2, position = position_dodge(width = 0.9)) + 
  labs(title = "Distribution of Wealth in Chosen Countries in Recent Year", x = "")

考察：