analysing number of parliament members worldwide

parliament_seats.R

require(dplyr)
require(rvest)
require(ggplot2)
require(ggthemes)

url <- "https://en.wikipedia.org/wiki/List_of_legislatures_by_number_of_members"
html_downloaded <- read_html(url)

tables<- url %>% read_html %>% 
  html_table(fill = T)

parliament_seats <- tables[[1]]

clean_text <- function(text){
  gsub(',', '', gsub('.*♠(.*)', '\\1', text))
}

parliament_seats <- parliament_seats %>%
  mutate_at(vars(`Lowerhouse[1]`:`Population/seats`), funs(clean_text)) %>%
  mutate_at(vars(`Lowerhouse[1]`:`Upperhouse[1]`, Total), funs(gsub('([0-9]*).*', '\\1', .))) %>%
  mutate_at(vars(`Lowerhouse[1]`:`Population/seats`), funs(as.numeric)) %>%
  replace_na(list(`Upperhouse[1]` = 0, `Lowerhouse[1]` = 0)) %>%
  mutate(Total = `Lowerhouse[1]`+`Upperhouse[1]`,
         `Population/seats`=`Population[2]`/Total)


ggplot(parliament_seats, aes(x=`Population[2]`, y=`Total`, 
                             color = ifelse(Type == 'bicameral', 'Dwuizbowy', 'Jednoizbowy'))) +
  geom_point() +
  scale_x_log10(labels=scales::number_format()) +
  scale_y_log10(labels=scales::number_format()) +
  scale_color_viridis_d() +
  theme_fivethirtyeight(base_family = 'Helvetica Neue Light') +
  theme(axis.title = element_text(inherit.blank = F)) +
  labs(x = NULL, y = 'Miejsc w parlamencie', color = NULL, title = 'Ludność kraju a liczba parlamentarzystów')

# fittin simple model
seats_pop_lm <- lm(I(log(`Total`))~log(`Population[2]`) + I(log(`Population[2]`)^2) + `Type`, 
                                                                  data = parliament_seats)
#quick diagnostics and summary
plot(seats_pop_lm, 1:2)
summary(seats_pop_lm)

#in Poland there is over representations
seats_pop_lm$residuals[parliament_seats$Country=='Poland']
#in reality it is 560
exp(seats_pop_lm$fitted.values)[parliament_seats$Country=='Poland']


parliament_seats %>%
  mutate(type = ifelse(Country == 'Poland', 'Polska', ifelse(Type == 'bicameral', 'Dwuizbowy parlament', 'Jednoizbowy parlament'))) %>%
  ggplot(aes(x=`Population[2]`, y=`Total`, color = type)) +
  geom_point() +
  scale_x_log10(labels=scales::number_format()) +
  scale_y_log10(labels=scales::number_format()) +
  scale_color_viridis_d() +
  facet_wrap(~cut(`Population[2]`, c(0, 1e6, 3e7, 1e10), labels = c('poniżej miliona', 'mniej niż 20 milionów', 'powyżej 10 milionów')),
             scales = 'free') +
  theme_fivethirtyeight(base_family = 'Helvetica Neue Light') +
  theme(axis.title = element_text(inherit.blank = F)) +
  labs(x = NULL, color = NULL, y = 'Mieszkańców na jedno miejsce w Parlamencie', title = 'Ludność kraju a liczba parlamentarzystów') +
  ggsave('miejsca_per_capita_populacja.png', width = 11, height = 7)


# add info about region from world bank metadata
# for example for https://data.worldbank.org/indicator/SP.POP.TOTL
metadata <- read.csv(METADATA_FILE, stringsAsFactors = F)

#adjust couple of non-matching names
parliament_seats$Country[parliament_seats$Country=="Korea, South"]="Korea, Rep."
parliament_seats$Country[parliament_seats$Country=="Korea, North"]="Korea, Dem. People's Rep."
parliament_seats$Country[parliament_seats$Country=="Iran"]="Iran, Islamic Rep."
parliament_seats$Country[parliament_seats$Country=="Russia"]="Russian Federation"
parliament_seats$Country[parliament_seats$Country=="Laos"]="Lao PDR"
parliament_seats$Country[parliament_seats$Country=="Venezuela"]="Venezuela, RB"
parliament_seats$Country[parliament_seats$Country=="Slovakia"]="Slovak Republic"

model_data <- parliament_seats %>%
  inner_join(metadata, by = c('Country' = 'TableName'))

#new model with additional variable - Region
seats_pop_lm2 <- lm(I(log(`Total`))~log(`Population[2]`) + I(log(`Population[2]`)^2) + `Type` + Region, 
                   data = model_data)
#we simplify model a bit
seats_pop_lm2 <- step(seats_pop_lm2, trace = F)
plot(seats_pop_lm)
summary(seats_pop_lm2)

seats_pop_lm2$residuals[model_data$Country=='Poland']
exp(seats_pop_lm2$fitted.values)[model_data$Country=='Poland']