R/viz_emphours_sex_recessions.R

In MattCowgill/macrodashboard: Shiny dashboard for Australian macroeconomic data

viz_emphours_sex_recessions <- function(data = load_data(),
                                        measure = "employment",
                                        facet_on = "recession" ) {
  stopifnot(measure %in% c("employment", "hours", "hours_pop"))
  stopifnot(facet_on %in% c("recession", "sex"))

  lfs_raw <- data$lfs_m_1
  hours_raw <- data$lfs_m_19

  # Define recession peaks ----
  # Based on emp-pop ratio for all persons
  tps <- dplyr::tribble(
    ~peak_date, ~min_date, ~recession,
    "1981-09-01", "1983-04-01", 1981L,
    "1990-05-01", "1993-02-01", 1989L,
    "2008-08-01", "2009-10-01", 2008L
  ) %>%
    mutate(across(ends_with("date"), as.Date)) %>%
    bind_rows(dplyr::tibble(
      peak_date = as.Date("2020-02-01"),
      min_date = max(lfs_raw$date),
      recession = 2020L
    ))


  dates_between_tps <- tps %>%
    group_by(.data$recession) %>%
    summarise(date = list(seq.Date(
      from = .data$peak_date,
      to = .data$min_date,
      by = "1 day"
    ))) %>%
    unnest(.data$date)

  # Define functions to get employment, hours, or hours per capita

  get_emp <- function(df = lfs_raw) {
    df %>%
      filter(.data$series_id %in% c(
        "A84423057T",
        "A84423029J"
      ) &
        !is.na(.data$value)) %>%
      separate_series() %>%
      mutate(series = "employed") %>%
      select(.data$date, .data$series, sex = .data$series_2, .data$value)
  }

  get_hours <- function(df = hours_raw) {
    df %>%
      filter(!is.na(.data$value) &
        .data$series_id %in% c(
          "A84426280L",
          "A84426274T"
        )) %>%
      separate_series() %>%
      mutate(series = "hours") %>%
      select(.data$date, .data$series, sex = .data$series_2, .data$value)
  }

  get_pop <- function(df = lfs_raw) {
    df %>%
      filter(!is.na(.data$value) &
        .data$series_id %in% c(
          "A84423105X",
          "A84423077A"
        )) %>%
      separate_series() %>%
      mutate(series = "pop") %>%
      select(.data$date, .data$series, sex = .data$series_2, .data$value)
  }

  get_hours_pop <- function() {
    hours <- get_hours()
    pop <- get_pop()
    df <- bind_rows(hours, pop)

    df %>%
      spread(key = .data$series, value = .data$value) %>%
      filter(!is.na(.data$hours)) %>%
      mutate(
        value = .data$hours / .data$pop,
        series = "hours_pop"
      ) %>%
      select(.data$date, .data$series, .data$sex, .data$value)
  }



  if (measure == "employment") {
    df <- get_emp()
  } else if (measure == "hours") {
    df <- get_hours()
  } else if (measure == "hours_pop") {
    df <- get_hours_pop()
  }


  # Get employment by sex during recessions
  recession_df <- df %>%
    left_join(dates_between_tps, by = "date") %>%
    filter(!is.na(.data$recession))

  recession_df <- recession_df %>%
    mutate(sex = case_when(
      .data$sex == "Females" ~ "Women",
      .data$sex == "Males" ~ "Men",
      TRUE ~ NA_character_
    )) %>%
    group_by(.data$recession, .data$sex) %>%
    mutate(
      change_since_peak = (.data$value / .data$value[date == min(.data$date)]) - 1,
      months_since_peak = row_number() - 1,
      recession_desc = case_when(
        .data$recession == 1981 ~ "1980s recession",
        .data$recession %in% c(1989, 1990) ~ "1990s recession",
        .data$recession == 2008 ~ "GFC",
        .data$recession == 2020 ~ "COVID"
      ),
      recession_desc = factor(.data$recession_desc,
        levels = c(
          "1980s recession",
          "1990s recession",
          "GFC",
          "COVID"
        ),
        ordered = T
      )
    )


  # Line chart -- emp change by sex during recessions
  title_text <- case_when(
    measure == "employment" ~ "employment",
    measure == "hours" ~ "total hours worked",
    measure == "hours_pop" ~ "hours per person"
  )

  facet_by_recession_plot <- function() {
    recession_df %>%
      ggplot(aes(x = .data$months_since_peak, y = .data$change_since_peak, col = .data$sex)) +
      geom_hline(yintercept = 0) +
      geom_line() +
      grattan_label_repel(
        data = ~ filter(., .data$date == max(.data$date)),
        aes(label = paste0(
          if_else(.data$recession == 1981,
            paste0(
              .data$sex,
              "\n"
            ),
            ""
          ),
          round(.data$change_since_peak * 100, 1), "%"
        )),
        nudge_x = 2,
        direction = "y",
        segment.size = 0,
        size = 14,
        hjust = 0
      ) +
      geom_point(
        data = ~ filter(., .data$date == max(.data$date)),
        stroke = 1.5,
        shape = 21,
        fill = "white"
      ) +
      facet_wrap(~ .data$recession_desc, nrow = 1) +
      scale_x_continuous(
        expand = expansion(add = c(1, 16)),
        breaks = seq(0, 36, 12),
        labels = function(x) x / 12
      ) +
      scale_y_continuous(labels = function(x) paste0(x * 100, "%")) +
      grattan_colour_manual(2) +
      theme_grattan() +
      labs(
        x = "Years since downturn started",
        captions = "Notes: Start of downturn defined as the month in which the total employment-to-population ratio reached its peak; trough is the month in which the ratio reached its lowest point. COVID recession starts at February 2020. Sources: ABS Labour Force and Grattan calculations.",
        title = paste0(tools::toTitleCase(title_text), " during Australian downturns"),
        subtitle = paste0("Change in ", title_text, " between start of downturn and trough")
      )
  }

  facet_by_sex_plot <- function() {
    recession_df %>%
      mutate(recession_desc = gsub(" recession", "", .data$recession_desc)) %>%
      ggplot(aes(x = .data$months_since_peak, y = .data$change_since_peak, col = .data$recession_desc)) +
      geom_hline(yintercept = 0) +
      geom_line() +
      grattan_label_repel(
        data = ~ filter(., .data$date == max(.data$date)),
        aes(label = paste0(
          stringr::str_wrap(.data$recession_desc, 10),
          "\n",
          round(.data$change_since_peak * 100, 1), "%"
        )),
        nudge_x = 1,
        direction = "y",
        segment.size = 0,
        size = 16,
        hjust = 0
      ) +
      geom_point(
        data = ~ filter(., .data$date == max(.data$date)),
        stroke = 1.5,
        shape = 21,
        fill = "white"
      ) +
      facet_wrap(~ .data$sex, nrow = 1) +
      scale_x_continuous(
        expand = expansion(add = c(1, 10)),
        breaks = seq(0, 36, 12),
        labels = function(x) x / 12
      ) +
      scale_y_continuous(labels = function(x) paste0(x * 100, "%")) +
      grattan_colour_manual() +
      theme_grattan() +
      labs(
        x = "Years since downturn started",
        captions = "Notes: Start of downturn defined as the month in which the total employment-to-population ratio reached its peak; trough is the month in which the ratio reached its lowest point. COVID recession starts at February 2020. Sources: ABS Labour Force and Grattan calculations.",
        title = paste0(tools::toTitleCase(title_text), " during Australian downturns"),
        subtitle = paste0("Change in ", title_text, " between start of downturn and trough")
      )
  }



  if (facet_on == "recession") {
    facet_by_recession_plot()
  } else {
    facet_by_sex_plot()
  }
}