Use case example: Vote Choice Over Time

Show code used in this page

library(dplyr)
library(ggplot2)
library(knitr)

pkg_master <- system.file("extdata", "qes_master.csv", package = "qesR")
master_paths <- c("qes_master.csv", "../qes_master.csv", pkg_master)
master_paths <- master_paths[nzchar(master_paths)]
master <- read.csv(master_paths[file.exists(master_paths)][1], stringsAsFactors = FALSE)

master <- master %>%
  mutate(
    qes_year_num = as.integer(substr(as.character(qes_year), 1, 4)),
    vote_choice_raw = trimws(as.character(vote_choice)),
    vote_choice_party = dplyr::case_when(
      vote_choice_raw %in% c("CAQ", "ADQ") ~ "CAQ/ADQ",
      vote_choice_raw == "PLQ" ~ "PLQ",
      vote_choice_raw == "PQ" ~ "PQ",
      vote_choice_raw == "QS" ~ "QS",
      vote_choice_raw == "PCQ" ~ "PCQ",
      TRUE ~ NA_character_
    )
  )

major_parties <- c("CAQ/ADQ", "PLQ", "PQ", "QS", "PCQ")
years <- sort(unique(master$qes_year_num[!is.na(master$qes_year_num)]))

vote_base <- master %>%
  filter(!is.na(qes_year_num), !is.na(vote_choice_party), nzchar(vote_choice_party))

vote_den <- vote_base %>%
  group_by(qes_year_num) %>%
  summarise(n_with_vote_choice = n(), .groups = "drop")

vote_party <- vote_base %>%
  group_by(qes_year_num, vote_choice_party) %>%
  summarise(n_party = n(), .groups = "drop")

grid <- expand.grid(
  qes_year_num = years,
  vote_choice_party = major_parties,
  stringsAsFactors = FALSE
)

vote_party <- merge(grid, vote_party, by = c("qes_year_num", "vote_choice_party"), all.x = TRUE, sort = TRUE)
vote_party$n_party[is.na(vote_party$n_party)] <- 0
vote_party <- merge(vote_party, vote_den, by = "qes_year_num", all.x = TRUE, sort = TRUE)

vote_party <- vote_party %>%
  mutate(
    share = ifelse(n_with_vote_choice > 0, n_party / n_with_vote_choice, NA_real_),
    se = ifelse(n_with_vote_choice > 0, sqrt(pmax(share * (1 - share), 0) / n_with_vote_choice), NA_real_),
    ci_low = ifelse(!is.na(se), pmax(0, share - 1.96 * se), NA_real_),
    ci_high = ifelse(!is.na(se), pmin(1, share + 1.96 * se), NA_real_),
    year_index = match(qes_year_num, years)
  )

vote_table <- vote_party %>%
  group_by(qes_year_num) %>%
  summarise(
    n_with_vote_choice = first(n_with_vote_choice),
    caq_adq = 100 * share[vote_choice_party == "CAQ/ADQ"],
    plq = 100 * share[vote_choice_party == "PLQ"],
    pq = 100 * share[vote_choice_party == "PQ"],
    qs = 100 * share[vote_choice_party == "QS"],
    pcq = 100 * share[vote_choice_party == "PCQ"],
    .groups = "drop"
  )

knitr::kable(vote_table)

Table 1: vote-choice shares by year

Major-party vote-choice shares by year

Study year	N with vote-choice item	CAQ/ADQ (%)	PLQ (%)	PQ (%)	QS (%)	PCQ (%)
1998	1048	26.4	31.0	42.6	0.0	0.0
2007	6852	33.6	27.7	33.9	4.9	0.0
2008	8902	19.3	38.6	36.4	5.6	0.0
2009	6121	10.4	40.6	41.2	7.7	0.0
2010	730	7.8	41.6	41.2	9.3	0.0
2012	1813	26.2	24.5	41.5	7.8	0.0
2014	1241	22.3	39.4	27.7	10.6	0.0
2018	2582	37.3	26.7	19.4	16.6	0.0
2022	1196	37.6	10.9	13.7	22.0	15.8

Figure 1: vote-choice trend

Notes

• Shares use respondents with a non-missing vote_choice value.
• Years 2009 and 2010 come from qes_crop_2007_2010 after splitting by collection year.
• Confidence intervals are binomial 95% intervals (normal approximation).
• The trend figure is restricted to CAQ/ADQ, PLQ, PQ, QS, and PCQ.