Use HILDA to calculate Gini by household

HILDA-household-disp-Gini.R

library(data.table)
library(hildaData)
library(magrittr)
library(grattanCharts)
library(ggplot2)

#' @source acid package (but it's chock full of unnecessary dependencies)
weighted_gini <- function (x, w = NULL) {
  if (is.null(w)) 
    w <- rep(1, length(x))
  x.sort <- sort(x)
  x.order <- order(x)
  x <- x.sort
  n <- length(x)
  w <- w[x.order]/sum(w)
  w.c <- cumsum(w)
  xw.c <- cumsum(w * x)
  xw.c <- xw.c/xw.c[n]
  Gini <- t(xw.c[-1]) %*% w.c[-n] - t(xw.c[-n]) %*% w.c[-1]
  Gini[1, 1]
}

get_disposable_incomes <- function(year) {
  wave <- letters[year - 2000]
  
  get(paste0("Combined_", wave, "150c")) %>%
    as.data.table %>%
    .[, .SD, .SDcols = c(#paste0(wave, "hhrhid"), 
                         paste0(wave, "hhwth"),
                         grep("hifdit[pn]", names(.), value = TRUE))] %>%
    setnames(paste0(wave, "hhwth"), "WEIGHT") %>%
    setnames(paste0(wave, "hifditp"), "positive_income") %>%
    setnames(paste0(wave, "hifditn"), "negative_income") %>%
    make_negatives_NA %>%
    .[, Disposable_Income := positive_income - negative_income] %>%
    .[, Year := year]
}

2001:2015 %>%
  lapply(get_disposable_incomes) %>%
  rbindlist %>%
  .[, .(Gini = weighted_gini(Disposable_Income, WEIGHT)), keyby = Year] %>%
  grplot(aes(x = Year, y = Gini)) + 
  geom_line()