Governance and accountability in Indian public opinion
World Values Survey 2001–2023: India, Western Europe & Offshoots, Middle East & North Africa
Arslan
06 October 2026
knitr::opts_chunk$set(message = FALSE, warning = FALSE)
options(knitr.table.format = "html") # needed for the <br>/<small> cell markup
library(dplyr)
library(tidyr)
library(purrr)
library(tibble)
library(ggplot2)
library(ggrepel)
library(knitr)
library(kableExtra)
library(survey)
library(forcats)
library(ggiraph)
# All cleaning and categorisation happens in R/00_wvs_clean_harmonise.R
wvs_all <- readRDS("D:/Populism and Democrary/World value survey/data/clean/wvs_harmonised.rds")
# Comparison set: India vs Western Europe & Offshoots and Middle East & North Africa.
# Adults only: W7 India deliberately interviewed 16-17 year olds (321 of 1,692).
focus_regions <- c("India", "Western Europe & Offshoots", "Middle East & North Africa")
# W3 (1994-99, India 1995) is dropped from this report
wvs <- wvs_all |>
filter(is.na(age) | age >= 18, macro_region %in% focus_regions, wave != "W3") |>
mutate(macro_region = droplevels(macro_region))
india_years <- c(W4 = "2001", W5 = "2006", W6 = "2012", W7 = "2023")
wave_labels <- c(W4 = "W4 (1999–2004)", W5 = "W5 (2004–09)", W6 = "W6 (2010–16)", W7 = "W7 (2017–23)")
common_states <- c("Bihar", "Delhi", "Haryana", "Maharashtra", "Punjab", "Uttar Pradesh", "West Bengal")
styled <- function(k) kable_styling(k, bootstrap_options = c("striped", "condensed"), full_width = FALSE)
# ---- Chart style ------------------------------------------------------------------
# Colour follows the group, in a fixed order (validated for colour-blind separation):
# Western Europe = blue, India = orange, MENA = green.
region_cols <- c(`Western Europe & Offshoots` = "#2a78d6", India = "#eb6834",
`Middle East & North Africa` = "#1baf7a")
year_cols <- c(`2001` = "#86b6ef", `2006` = "#5598e7",
`2012` = "#256abf", `2023` = "#0d366b") # one-hue ramp: later = darker
india_year_num <- c(W4 = 2001, W5 = 2006, W6 = 2012, W7 = 2023)
ink <- "#0b0b0b"; ink2 <- "#52514e"; grid_col <- "#e6e5e1"
theme_wvs <- function(base = 11) {
theme_minimal(base_size = base) +
theme(panel.grid.minor = element_blank(),
panel.grid.major = element_line(colour = grid_col, linewidth = 0.3),
axis.text = element_text(colour = ink2), axis.title = element_text(colour = ink2),
plot.title = element_text(face = "bold", size = base + 1, colour = ink),
plot.subtitle = element_text(colour = ink2),
plot.caption = element_text(colour = ink2, hjust = 0, size = base - 2),
legend.position = "top", legend.title = element_blank(), legend.justification = "left",
strip.text = element_text(face = "bold", hjust = 0, colour = ink))
}
knitr::opts_chunk$set(dpi = 110)
# Every chart is rendered with ggiraph: hover a dot, bar or point to see its value
# ggiraph tooltips / ids cannot contain a straight apostrophe
safe_txt <- function(x) gsub("'", "’", x, fixed = TRUE)
show_plot <- function(p, w = 9, h = 5) {
if (isTRUE(getOption("wvs.static"))) return(p) # for checking layout as images
girafe(ggobj = p, width_svg = w, height_svg = h,
options = list(
opts_tooltip(css = paste0("background:#ffffff;color:#0b0b0b;border:1px solid #d9d8d4;",
"border-radius:4px;padding:6px 8px;font-size:12px;")),
opts_hover(css = "stroke:#0b0b0b;stroke-width:1.5px;"),
opts_hover_inv(css = "opacity:0.4;"),
opts_sizing(rescale = TRUE),
opts_toolbar(saveaspng = TRUE)))
}agree4 <- function(x) case_when(x %in% 1:2 ~ 1, x %in% 3:4 ~ 0)
agree5 <- function(x) case_when(x %in% 1:2 ~ 1, x %in% 3:5 ~ 0)
high10 <- function(x) case_when(x %in% 7:10 ~ 1, x %in% 1:6 ~ 0) # top 4 points of a 1-10 scale
# mean of 0/1 items when at least `k` are answered
row_share <- function(d, vars, k) {
n_ok <- rowSums(!is.na(d[vars]))
ifelse(n_ok >= k, rowMeans(d[vars], na.rm = TRUE), NA_real_)
}
# Institutions that USE power (the state) vs institutions that CHECK power
state_inst <- c("conf_civilservice_hi", "conf_police_hi", "conf_army_hi", "conf_govt_hi")
check_inst <- c("conf_parliament_hi", "conf_parties_hi", "conf_press_hi", "conf_courts_hi")
conc_levels <- c("Liberal (maximalist)", "Minimalist (electoral)", "Illiberal")
add_recodes <- function(d) {
d |>
mutate(
# 1. trust in the two sides of the state (share of 4 institutions trusted, 0-1)
state_trust = row_share(pick(everything()), state_inst, 3),
check_trust = row_share(pick(everything()), check_inst, 3),
trust_gap = state_trust - check_trust,
# 2. demand for governance (delivery) and for accountability: yes/no on each question (combined below)
# governance = what people want government to deliver (ranked aims: picked 1st or 2nd)
gov_growth = case_when(pm_aims1 == 1 | pm_aims2 == 1 ~ 1, pm_aims1 %in% 1:4 ~ 0),
gov_prices = case_when(pm_goals1 == 3 | pm_goals2 == 3 ~ 1, pm_goals1 %in% 1:4 ~ 0),
gov_economy = case_when(pm_society1 == 1 | pm_society2 == 1 ~ 1, pm_society1 %in% 1:4 ~ 0),
gov_crime = case_when(pm_society1 == 4 | pm_society2 == 4 ~ 1, pm_society1 %in% 1:4 ~ 0),
acc_say_govt = case_when(pm_goals1 == 2 | pm_goals2 == 2 ~ 1, pm_goals1 %in% 1:4 ~ 0),
acc_say_work = case_when(pm_aims1 == 3 | pm_aims2 == 3 ~ 1, pm_aims1 %in% 1:4 ~ 0),
acc_petition = case_when(act_petition %in% 1:2 ~ 1, act_petition == 3 ~ 0),
acc_demo = case_when(act_demonstration %in% 1:2 ~ 1, act_demonstration == 3 ~ 0),
acc_boycott = case_when(act_boycott %in% 1:2 ~ 1, act_boycott == 3 ~ 0),
# 3. conception of democracy: yes/no on each liberal and each authoritarian question
lib_elections = case_when(dess_elections %in% 7:10 ~ 1, dess_elections %in% 1:6 ~ 0),
lib_civil = case_when(dess_civilrights %in% 7:10 ~ 1, dess_civilrights %in% 1:6 ~ 0),
lib_women = case_when(dess_women %in% 7:10 ~ 1, dess_women %in% 1:6 ~ 0),
lib_demgood = case_when(reg_democracy %in% 1:2 ~ 1, reg_democracy %in% 3:4 ~ 0),
lib_demimp = case_when(dem_importance %in% 7:10 ~ 1, dem_importance %in% 1:6 ~ 0),
aut_army_ess = case_when(dess_army %in% 7:10 ~ 1, dess_army %in% 1:6 ~ 0),
aut_relig_ess = case_when(dess_religious %in% 7:10 ~ 1, dess_religious %in% 1:6 ~ 0),
aut_army_rule = case_when(reg_army %in% 1:2 ~ 1, reg_army %in% 3:4 ~ 0),
aut_strong = case_when(reg_strongleader %in% 1:2 ~ 1, reg_strongleader %in% 3:4 ~ 0),
aut_experts = case_when(reg_experts %in% 1:2 ~ 1, reg_experts %in% 3:4 ~ 0),
aut_obey = case_when(dess_obey %in% 7:10 ~ 1, dess_obey %in% 1:6 ~ 0),
aut_authority = case_when(fut_authority == 1 ~ 1, fut_authority %in% 2:3 ~ 0),
# liberal score and authoritarian score = share of answered questions with a "yes" (0-1);
# democracy score = liberal score minus authoritarian score (-1 to +1)
aut_share = rowMeans(pick(aut_army_ess, aut_relig_ess, aut_army_rule, aut_strong,
aut_experts, aut_obey, aut_authority), na.rm = TRUE),
lib_share = rowMeans(pick(lib_elections, lib_civil, lib_women, lib_demgood, lib_demimp), na.rm = TRUE),
dem_score = lib_share - aut_share,
# Classified if free elections was answered (asked from 2006) and at least 3 questions in
# each set were answered.
aut_n = rowSums(!is.na(pick(aut_army_ess, aut_relig_ess, aut_army_rule, aut_strong,
aut_experts, aut_obey, aut_authority))),
lib_n = rowSums(!is.na(pick(lib_elections, lib_civil, lib_women, lib_demgood, lib_demimp))),
ess_ok = !is.na(dess_elections) & aut_n >= 3 & lib_n >= 3,
# Types are bands of the democracy score:
# Illiberal: score <= 0, or free elections rated 1-6
# Minimalist: free elections 7-10 and score above 0 but below 0.5
# Liberal: free elections 7-10 and score 0.5 or more
conception = case_when(
!ess_ok ~ NA_character_,
lib_elections == 0 | dem_score <= 0 ~ "Illiberal",
lib_elections == 1 & dem_score >= 0.5 ~ "Liberal (maximalist)",
lib_elections == 1 & dem_score > 0 & dem_score < 0.5 ~ "Minimalist (electoral)"),
conception = factor(conception, conc_levels),
con_liberal = as.numeric(conception == "Liberal (maximalist)"),
con_minimal = as.numeric(conception == "Minimalist (electoral)"),
con_illiberal = as.numeric(conception == "Illiberal"),
# state support (standalone dimension, section 4): yes on each item
ss_govresp_yes = case_when(econ_govresp %in% 1:4 ~ 1, econ_govresp %in% 5:10 ~ 0),
ss_income_yes = case_when(econ_income %in% 1:4 ~ 1, econ_income %in% 5:10 ~ 0),
ss_taxrich_yes = high10(dess_taxrich),
ss_unemp_yes = high10(dess_unemp),
ss_incomeeq_yes = high10(dess_incomeequal),
ss_policy_n = ss_govresp_yes + ss_income_yes,
ss_essential_n = ss_taxrich_yes + ss_unemp_yes + if_else(wave == "W5", 0, ss_incomeeq_yes)
)
}
wvs <- add_recodes(wvs)
# State supporter = yes on at least 3 of the state-support items. Classified only where at least
# 4 of the 5 were asked in that country-wave (so from 2006) and the respondent answered all asked.
ss_items <- c("ss_govresp_yes", "ss_income_yes", "ss_taxrich_yes", "ss_unemp_yes", "ss_incomeeq_yes")
wvs <- wvs |>
group_by(wave, country_code) |>
mutate(ss_asked = sum(map_lgl(pick(all_of(ss_items)), ~ any(!is.na(.x))))) |>
ungroup() |>
mutate(ss_answered = rowSums(!is.na(pick(all_of(ss_items)))),
ss_yes = rowSums(pick(all_of(ss_items)), na.rm = TRUE),
state_support = case_when(ss_asked < 4 | ss_answered < ss_asked ~ NA_real_,
ss_yes >= 3 ~ 1, TRUE ~ 0))
# Demand for governance / accountability = yes on at least half of the questions asked in that
# country-wave. Classified if the respondent answered all but at most one of those asked.
gov_items <- c("gov_growth", "gov_prices", "gov_economy", "gov_crime")
acc_items <- c("acc_say_govt", "acc_say_work", "acc_petition", "acc_demo", "acc_boycott")
demand <- function(d, items) {
asked <- d |>
group_by(wave, country_code) |>
summarise(k = sum(map_lgl(pick(all_of(items)), ~ any(!is.na(.x)))), .groups = "drop")
k <- left_join(select(d, wave, country_code), asked, by = c("wave", "country_code"))$k
ans <- rowSums(!is.na(d[items]))
yes <- rowSums(d[items], na.rm = TRUE)
ifelse(k >= 2 & ans >= k - 1, as.numeric(yes / ans >= 0.5), NA_real_)
}
wvs <- wvs |>
mutate(gov_demand = demand(wvs, gov_items),
acc_demand = demand(wvs, acc_items),
gtype = case_when(
acc_demand == 1 & gov_demand == 0 ~ "Accountability-first",
acc_demand == 1 & gov_demand == 1 ~ "Want both",
acc_demand == 0 & gov_demand == 1 ~ "Governance-first",
acc_demand == 0 & gov_demand == 0 ~ "Neither"),
gtype = factor(gtype, c("Accountability-first", "Want both", "Governance-first", "Neither")),
g_accfirst = as.numeric(gtype == "Accountability-first"),
g_both = as.numeric(gtype == "Want both"),
g_govfirst = as.numeric(gtype == "Governance-first"),
# robustness: accountability from stated preferences only (no demonstrations / boycotts)
acc_demand_pref = demand(wvs, setdiff(acc_items, c("acc_demo", "acc_boycott"))),
gtype_pref = case_when(
acc_demand_pref == 1 & gov_demand == 0 ~ "Accountability-first",
acc_demand_pref == 1 & gov_demand == 1 ~ "Want both",
acc_demand_pref == 0 & gov_demand == 1 ~ "Governance-first",
acc_demand_pref == 0 & gov_demand == 0 ~ "Neither"))
india <- wvs |> filter(is_india)# ---- Weighted estimate with 95% CI ------------------------------------------
# SEs use the Kish effective sample size (weights only). The WVS does not release
# sampling clusters, so intervals are too narrow; do not over-read small gaps.
wstat <- function(x, w) {
ok <- !is.na(x) & !is.na(w); x <- x[ok]; w <- w[ok]
if (!length(x)) return(tibble(est = NA_real_, se = NA_real_, n = 0L))
m <- sum(w * x) / sum(w)
neff <- sum(w)^2 / sum(w^2)
tibble(est = m, se = sqrt(sum(w * (x - m)^2) / sum(w) / neff), n = length(x))
}
wdk <- function(x, w) if (all(is.na(x))) NA_real_ else 100 * sum(w[is.na(x)]) / sum(w)
# ---- India over time ------------------------------------------------------------
india_trend <- function(items, mult = 100, data = india) {
imap_dfr(items, function(lab, v) {
data |>
group_by(wave) |>
group_modify(~ bind_cols(wstat(.x[[v]], .x$weight), dk = wdk(.x[[v]], .x$weight))) |>
ungroup() |>
mutate(item = lab, var = v)
}) |>
mutate(lo = est - 1.96 * se, hi = est + 1.96 * se)
}
show_trend <- function(tr, mult = 100, caption = NULL, digits = 1) {
f <- paste0("%.", digits, "f")
tr |>
mutate(cell = ifelse(is.na(est), "—",
sprintf(paste0(f, " [", f, ", ", f, "]<br><small>no answer %.0f%%</small>"),
est * mult, lo * mult, hi * mult, dk)),
year = india_years[wave]) |>
select(Item = item, year, cell) |>
pivot_wider(names_from = year, values_from = cell) |>
(\(x) kable(x, escape = FALSE, caption = caption, align = c("l", rep("c", ncol(x) - 1))))() |>
styled()
}
# ---- Country estimates and India vs regions -------------------------------------
# Region value = unweighted mean of country estimates (each country counts once)
country_est <- function(v, mult = 100, data = wvs) {
data |>
group_by(wave, country_code, country_name, macro_region) |>
filter(any(!is.na(.data[[v]]))) |>
summarise(est = weighted.mean(.data[[v]], weight, na.rm = TRUE) * mult, .groups = "drop")
}
region_compare <- function(v, mult = 100, caption = NULL, digits = 1) {
ce <- country_est(v, mult)
f <- paste0("%.", digits, "f")
reg <- ce |>
group_by(wave, macro_region) |>
summarise(cell = sprintf(paste0(f, " <small>(%d)</small>"), mean(est), n()), .groups = "drop") |>
mutate(macro_region = as.character(macro_region))
rnk <- ce |>
group_by(wave) |>
mutate(r = rank(-est, ties.method = "min"), N = n()) |>
filter(country_code == 356) |>
transmute(wave, macro_region = "India's rank among countries shown (1 = highest)",
cell = sprintf("%d of %d", r, N))
bind_rows(reg, rnk) |>
mutate(wave = wave_labels[wave]) |>
pivot_wider(names_from = wave, values_from = cell, values_fill = "—") |>
arrange(factor(macro_region, levels = c(focus_regions, "India's rank among countries shown (1 = highest)"))) |>
rename(Region = macro_region) |>
(\(x) kable(x, escape = FALSE, caption = caption, align = c("l", rep("c", ncol(x) - 1))))() |>
styled() |>
footnote(general = paste0("Cells: mean of country estimates (number of countries).",
if (grepl("trust|check", v)) " W4 has no courts item, so its checking-trust index uses 3 institutions." else ""),
general_title = "")
}
# ---- Charts used across sections --------------------------------------------------
# India over time: one or more items, India only (orange), items told apart by line type + labels
plot_india_lines <- function(tr, title, ylab = "%", mult = 100, caption = NULL, w = 9, h = 4.6) {
d <- tr |> filter(!is.na(est)) |> mutate(year = india_year_num[wave])
p <- ggplot(d, aes(year, est * mult, group = item)) +
geom_ribbon(aes(ymin = lo * mult, ymax = hi * mult), fill = region_cols["India"], alpha = 0.10) +
geom_line(aes(linetype = item), colour = region_cols["India"], linewidth = 0.8) +
geom_point_interactive(aes(tooltip = sprintf("%s<br>%d: %.1f (95%% CI %.1f to %.1f)", item, year,
est * mult, lo * mult, hi * mult),
data_id = paste(item, year)),
shape = 21, fill = region_cols["India"], colour = "white", size = 3, stroke = 0.8) +
geom_text(data = d |> group_by(item) |> slice_max(year, n = 1),
aes(label = item), hjust = -0.08, size = 3.1, colour = ink) +
scale_x_continuous(breaks = unname(india_year_num), limits = c(2000, 2033)) +
scale_linetype_manual(values = c("solid", "22", "42", "12")) +
labs(title = title, x = NULL, y = ylab, caption = caption) +
theme_wvs() + theme(legend.position = "none")
show_plot(p, w, h)
}
# India vs the two regions over waves: region = mean of countries; India = its own estimate
plot_region_lines <- function(v, title, ylab = "%", mult = 100, caption = NULL, w = 9, h = 4.6) {
d <- country_est(v, mult) |>
group_by(wave, macro_region) |>
summarise(est = mean(est), k = n(), .groups = "drop") |>
mutate(wave_lab = factor(wave_labels[wave], wave_labels), macro_region = factor(macro_region, focus_regions))
p <- ggplot(d, aes(wave_lab, est, colour = macro_region, group = macro_region)) +
geom_line(linewidth = 0.8) +
geom_point_interactive(aes(fill = macro_region, data_id = paste(macro_region, wave),
tooltip = sprintf("%s<br>%s: %.1f<br>%d %s", macro_region, wave_lab, est, k,
ifelse(k == 1, "country", "countries"))),
shape = 21, colour = "white", size = 3.2, stroke = 0.8) +
geom_text(data = d |> group_by(macro_region) |> slice_max(as.integer(wave_lab), n = 1),
aes(label = sprintf("%.0f", est)), hjust = -0.6, size = 3.1, colour = ink, show.legend = FALSE) +
scale_colour_manual(values = region_cols, breaks = focus_regions) +
scale_fill_manual(values = region_cols, guide = "none") +
labs(title = title, x = NULL, y = ylab,
caption = caption %||% "Regions: mean of country estimates. W. Europe in W4 is 4 countries.") +
theme_wvs()
show_plot(p, w, h)
}
# Every country as a dot, ranked within each wave (one panel per wave); India highlighted.
# Dashed lines = region averages (mean of countries).
plot_country_dots <- function(v, title, xlab, mult = 100, waves = NULL, caption = NULL, w = 11, h = NULL) {
ce <- country_est(v, mult)
if (!is.null(waves)) ce <- filter(ce, wave %in% waves)
ce <- ce |>
mutate(wave_lab = factor(wave_labels[wave], wave_labels),
india = country_code == 356,
name = safe_txt(as.character(country_name)),
key = paste(name, wave, sep = "___")) |>
arrange(wave, est) |>
mutate(key = factor(key, unique(key)))
rm <- ce |> filter(!india) |> group_by(wave_lab, macro_region) |> summarise(m = mean(est), .groups = "drop")
n_max <- max(table(ce$wave))
p <- ggplot(ce, aes(est, key)) +
geom_vline(data = rm, aes(xintercept = m, colour = macro_region), linetype = "22", linewidth = 0.5) +
geom_point_interactive(aes(colour = macro_region, size = india, data_id = paste(name, wave),
tooltip = sprintf("%s (%s)<br>%s: %.1f", name, macro_region, wave_lab, est))) +
facet_wrap(~ wave_lab, nrow = 1, scales = "free_y") +
scale_y_discrete(labels = function(x) sub("___.*", "", x)) +
scale_colour_manual(values = region_cols, breaks = focus_regions) +
scale_size_manual(values = c(`FALSE` = 2.2, `TRUE` = 3.8), guide = "none") +
labs(title = title, x = xlab, y = NULL,
caption = caption %||% "Countries ranked within each wave. Dashed lines: region averages (mean of countries).") +
theme_wvs(10) + theme(panel.spacing = unit(1.4, "lines"), panel.grid.major.y = element_blank(),
axis.text.y = element_text(size = 7.5))
show_plot(p, w, h %||% max(4.5, 1.8 + 0.18 * n_max))
}
# ---- Groups within India ------------------------------------------------------------
india_groups <- function(v, mult = 100, min_n = 50, caption = NULL, digits = 1,
groups = c(Sex = "sex_f", Age = "age_cat3", Education = "educ3",
Party = "in_party_vote", Religion = "in_religion", Caste = "in_caste")) {
keep_levels <- list(in_party_vote = c("BJP", "INC", "Left", "Other party"),
in_religion = c("Hindu", "Muslim", "Christian", "Other (Sikh, Buddhist, Jain, ...)"))
f <- paste0("%.", digits, "f")
long <- imap_dfr(groups, function(g, gl) {
d <- india |> filter(!is.na(.data[[g]]))
if (g %in% names(keep_levels)) d <- d |> filter(.data[[g]] %in% keep_levels[[g]])
d |>
mutate(level = factor(as.character(.data[[g]]), levels = levels(droplevels(factor(.data[[g]]))))) |>
group_by(wave, level) |>
group_modify(~ wstat(.x[[v]], .x$weight)) |>
ungroup() |>
mutate(group = factor(gl, levels = names(groups)))
}) |>
filter(n > 0)
wide <- long |>
mutate(cell = ifelse(n < min_n, sprintf("<small>n=%d</small>", n),
sprintf(paste0(f, " <small>(%d)</small>"), est * mult, n)),
year = factor(india_years[wave], levels = india_years)) |>
arrange(group, level, year) |>
select(group, level, year, cell) |>
pivot_wider(names_from = year, values_from = cell, values_fill = "—")
wide |>
select(-group) |>
rename(` ` = level) |>
kable(escape = FALSE, caption = caption, align = c("l", rep("c", ncol(wide) - 2))) |>
styled() |>
pack_rows(index = table(droplevels(wide$group))) |>
footnote(general = sprintf("Cells: estimate (unweighted n). Cells with n < %d are not estimated.", min_n),
general_title = "")
}
# Hindu vs Muslim over time, one panel per item
relig_cols <- c(Hindu = "#4a3aa7", Muslim = "#eda100")
plot_relig_lines <- function(items, title, ylab = "%", ncol = 2, w = 10, h = 6.5, caption = NULL) {
d <- imap_dfr(items, function(lab, v) {
india |>
filter(in_religion %in% names(relig_cols)) |>
group_by(wave, in_religion) |>
group_modify(~ wstat(.x[[v]], .x$weight)) |>
ungroup() |>
filter(n >= 50) |>
mutate(item = lab)
}) |>
mutate(year = india_year_num[wave], item = factor(item, items),
in_religion = factor(as.character(in_religion), names(relig_cols)))
p <- ggplot(d, aes(year, est * 100, colour = in_religion, group = in_religion)) +
geom_line(linewidth = 0.8) +
geom_point_interactive(aes(fill = in_religion, data_id = safe_txt(paste(item, in_religion, year)),
tooltip = safe_txt(sprintf("%s<br>%s, %d: %.1f%%<br>n = %d", item, in_religion, year, est * 100, n))),
shape = 21, colour = "white", size = 2.8, stroke = 0.7) +
geom_text_repel(data = d |> group_by(item, in_religion) |> slice_max(year, n = 1),
aes(label = in_religion), size = 2.9, colour = ink, show.legend = FALSE,
direction = "y", hjust = 0, nudge_x = 1.2, segment.colour = NA, seed = 1) +
facet_wrap(~ item, ncol = ncol) +
scale_colour_manual(values = relig_cols) +
scale_fill_manual(values = relig_cols, guide = "none") +
scale_x_continuous(breaks = unname(india_year_num), limits = c(2000, 2028)) +
scale_y_continuous(limits = c(0, 100)) +
labs(title = title, x = NULL, y = ylab,
caption = caption %||% "India only. Groups with fewer than 50 respondents are not shown.") +
theme_wvs(10) + theme(panel.spacing = unit(1.2, "lines"))
show_plot(p, w, h)
}
# Several items in one wave: India and the two regions side by side (Cleveland dot plot)
plot_items_regions <- function(items, wave_sel, title, xlab = "%", w = 9.5, h = NULL, caption = NULL) {
d <- imap_dfr(items, function(lab, v) {
country_est(v) |>
filter(wave == wave_sel) |>
group_by(macro_region) |>
summarise(est = mean(est), k = n(), .groups = "drop") |>
mutate(item = lab)
}) |>
mutate(item = factor(item, rev(items)), macro_region = factor(macro_region, focus_regions))
p <- ggplot(d, aes(est, item)) +
geom_line(aes(group = item), colour = "#c3c2b7", linewidth = 0.9) +
geom_point_interactive(aes(colour = macro_region, data_id = safe_txt(paste(item, macro_region)),
tooltip = safe_txt(sprintf("%s<br>%s: %.0f%%<br>%d %s", item, macro_region, est, k,
ifelse(k == 1, "country", "countries")))),
size = 3.6) +
scale_colour_manual(values = region_cols, breaks = focus_regions) +
scale_x_continuous(limits = c(0, 100)) +
labs(title = title, x = xlab, y = NULL,
caption = caption %||% paste0(wave_labels[wave_sel], ". Regions: mean of countries.")) +
theme_wvs()
show_plot(p, w, h %||% (1.6 + 0.55 * length(items)))
}
# long version of the group estimates, for charts
group_long <- function(v, groups = c(Sex = "sex_f", Age = "age_cat3", Education = "educ3",
Party = "in_party_vote", Religion = "in_religion", Caste = "in_caste")) {
keep_levels <- list(in_party_vote = c("BJP", "INC", "Left", "Other party"),
in_religion = c("Hindu", "Muslim", "Christian", "Other (Sikh, Buddhist, Jain, ...)"))
imap_dfr(groups, function(g, gl) {
d <- india |> filter(!is.na(.data[[g]]))
if (g %in% names(keep_levels)) d <- d |> filter(.data[[g]] %in% keep_levels[[g]])
d |>
mutate(level = factor(as.character(.data[[g]]), levels = levels(droplevels(factor(.data[[g]]))))) |>
group_by(wave, level) |>
group_modify(~ wstat(.x[[v]], .x$weight)) |>
ungroup() |>
mutate(group = factor(gl, levels = names(groups)))
})
}
# ---- Question table at the start of each section ------------------------------------
fmt_conf <- "4-point: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all"
fmt_reg <- "4-point: 1 very good, 2 fairly good, 3 fairly bad, 4 very bad"
fmt_ess <- "10-point: 1 not an essential characteristic of democracy ... 10 an essential characteristic"
fmt_agr4 <- "4-point: 1 strongly agree, 2 agree, 3 disagree, 4 strongly disagree"
fmt_agr5 <- "5-point: 1 agree strongly ... 3 neither ... 5 disagree strongly"
fmt_just <- "10-point: 1 never justifiable ... 10 always justifiable"
fmt_surv <- "4-point: 1 definitely should ... 4 definitely should not"
fmt_nb <- "Yes/no: mentioned or not mentioned on a list"
codebook_vars <- readr::read_csv("D:/Populism and Democrary/World value survey/data/clean/wvs_codebook.csv",
show_col_types = FALSE) |>
select(var, W4, W5, W6, W7)
# rows: var (several joined with "+", e.g. "pm_aims1+pm_aims2"), question, format, mapped,
# and optionally `side` to group rows under sub-headings
q_table <- function(rows, caption = "Questions used in this section") {
first <- sub("\\+.*", "", rows$var)
in_india <- vapply(first, function(v) {
yrs <- india |> filter(!is.na(.data[[v]])) |> distinct(wave) |> pull(wave) |> sort()
if (length(yrs)) paste(india_years[yrs], collapse = ", ") else "—"
}, character(1))
wave_ids <- function(v, w) {
ids <- strsplit(v, "+", fixed = TRUE)[[1]]
x <- codebook_vars[[w]][match(ids, codebook_vars$var)]
x <- x[!is.na(x) & x != ""]
if (length(x)) paste(x, collapse = " & ") else "—"
}
tab <- rows |> mutate(india_yrs = unname(in_india))
for (w in c("W4", "W5", "W6", "W7")) tab[[w]] <- vapply(rows$var, wave_ids, character(1), w = w)
k <- tab |>
select(Question = question, `Response format` = format, `How it is counted / what it maps to` = mapped,
`Asked in India` = india_yrs, W4, W5, W6, W7) |>
kable(caption = caption) |>
styled() |>
add_header_above(c(" " = 4, "Question number in the WVS file" = 4)) |>
column_spec(1, width = "17em") |>
column_spec(3, width = "15em")
if ("side" %in% names(rows)) k <- pack_rows(k, index = table(factor(rows$side, unique(rows$side))))
k
}
# India over time, one number per cell (no intervals)
simple_trend <- function(tr, caption, mult = 100, digits = 0, footnote_text = NULL) {
k <- tr |>
mutate(cell = ifelse(is.na(est), "—", formatC(est * mult, format = "f", digits = digits)),
year = india_years[wave]) |>
select(item, year, cell) |>
pivot_wider(names_from = year, values_from = cell) |>
rename(` ` = item) |>
(\(x) kable(x, caption = caption, align = c("l", rep("c", ncol(x) - 1))))() |>
styled()
if (!is.null(footnote_text)) k <- footnote(k, general = footnote_text, general_title = "")
k
}
# change within India by group: one dot per survey year, joined by a line
plot_dumbbell <- function(v, waves, title, xlab, min_n = 50, mult = 100, w = 9, h = 7.5) {
gl <- group_long(v) |>
filter(wave %in% waves, n >= min_n) |>
mutate(year = factor(india_years[wave], india_years[waves]), level = fct_rev(level))
p <- ggplot(gl, aes(est * mult, level)) +
geom_line(aes(group = level), colour = "#c3c2b7", linewidth = 0.9) +
geom_point_interactive(aes(fill = year, data_id = paste(group, level, year),
tooltip = sprintf("%s: %s<br>%s: %.1f%% (n = %d)", group, level, year, est * mult, n)),
shape = 21, colour = "white", size = 3.6, stroke = 0.8) +
facet_grid(group ~ ., scales = "free_y", space = "free_y", switch = "y") +
scale_fill_manual(values = year_cols) +
labs(title = title, x = xlab, y = NULL, caption = sprintf("Groups with fewer than %d respondents are not shown.", min_n)) +
theme_wvs() +
theme(strip.placement = "outside", strip.text.y.left = element_text(angle = 0, face = "bold"))
show_plot(p, w, h)
}What this report does
Part I sets out the data: which countries and waves, how India was sampled, who was interviewed, and how region averages are formed.
Part II asks one question from several angles: do Indians separate governance (a state that gets things done) from democratic accountability (checks on that state), and do they trade one for the other? The distinction between institutions that use power and institutions that check it comes from Fukuyama (2013, What Is Governance?). It is used here as a set of dimensions, not as the framework.
| # | Question | What is calculated |
|---|---|---|
| 1 | How much do people trust the state, and the institutions that check it? | Confidence in 8 institutions; a trust gap between the state and the institutions that check it |
| 2 | Do people want a state that delivers, a state that answers to them, or both? | Two yes/no demands (delivery; accountability, including willingness to act) crossed into four groups |
| 3 | Do people hold a liberal, a minimalist or an illiberal idea of democracy? | Three conception types; willingness to protest by type |
| 4 | Do people want the state to support the less well-off? | A yes/no state-support category from five questions |
| 5 | Does it differ across Indian states? | The main measures of sections 1 to 3 by state |
Comparison set: India against Western Europe & Offshoots and the Middle East & North Africa (MENA), for the four waves from W4 (India 2001) to W7 (India 2023). The earlier 1995 wave is not used. Western Europe has only 4 countries in W4 (Canada, Spain, Sweden, USA).
Two caveats on India’s samples, set out in Part I:
- The 2001, 2006 and 2012 samples carry no weights: they are male-skewed and, in 2006 and 2012, mostly rural.
- The 2023 sample is a quota sample in 8 states.
Part I. Data, samples and coverage
knitr::opts_chunk$set(message = FALSE, warning = FALSE)
options(knitr.table.format = "html")
library(dplyr)
library(tidyr)
library(purrr)
library(tibble)
library(knitr)
library(kableExtra)
# All cleaning and categorisation happens in R/00_wvs_clean_harmonise.R
wvs_all <- readRDS("D:/Populism and Democrary/World value survey/data/clean/wvs_harmonised.rds")
# Comparison set for this project: India vs two regions
focus_regions <- c("India", "Western Europe & Offshoots", "Middle East & North Africa")
# W3 (1994-99, India 1995) is dropped from this report
wvs <- wvs_all |>
filter(macro_region %in% focus_regions, wave != "W3") |>
mutate(macro_region = droplevels(macro_region))
wave_labels <- c(W4 = "W4", W5 = "W5", W6 = "W6", W7 = "W7")
india_years <- c(W4 = "2001", W5 = "2006", W6 = "2012", W7 = "2023")
regions <- levels(wvs$macro_region)
wmean <- function(x, w) {
ok <- !is.na(x) & !is.na(w)
if (!any(ok)) NA_real_ else sum(x[ok] * w[ok]) / sum(w[ok])
}
styled <- function(k) kable_styling(k, bootstrap_options = c("striped", "condensed"), full_width = FALSE)The comparison set is India against two regions: Western Europe & Offshoots and the Middle East & North Africa (MENA). Every table in this part is restricted to these three groups.
Before testing any claim, this part sets out:
- what each wave covers
- how countries are grouped into regions, and how region averages are weighted
- the demographic make-up of each region and of India
- how the India samples were drawn and weighted
- how often people give no answer
Part II uses the same data
(data/clean/wvs_harmonised.rds) and the same region
scheme.
The four waves
per_country <- wvs |> count(wave, country_code)
wvs |>
group_by(wave) |>
summarise(Fieldwork = paste(min(survey_year, na.rm = TRUE), max(survey_year, na.rm = TRUE), sep = "–"),
Countries = n_distinct(country_code),
Respondents = n(),
`Under 18` = sum(age < 18, na.rm = TRUE),
.groups = "drop") |>
left_join(per_country |> group_by(wave) |>
summarise(`Median N per country` = median(n), `Smallest` = min(n), `Largest` = max(n)),
by = "wave") |>
mutate(`India survey` = india_years[wave]) |>
rename(Wave = wave) |>
kable(caption = "Coverage of the four waves (India, Western Europe & Offshoots, MENA)", format.args = list(big.mark = ",")) |>
styled() |>
footnote(general = "W3 (1994–99) is not used. 'Under 18': respondents aged 15–17 (a few countries sample from 15 or 16); Part II uses adults only.",
general_title = "")| Wave | Fieldwork | Countries | Respondents | Under 18 | Median N per country | Smallest | Largest | India survey |
|---|---|---|---|---|---|---|---|---|
| W4 | 1999–2004 | 13 | 23,873 | 299 | 1,502.0 | 1,015 | 3,401 | 2001 |
| W5 | 2004–2008 | 23 | 33,658 | 107 | 1,200.0 | 954 | 3,051 | 2006 |
| W6 | 2010–2014 | 22 | 32,798 | 5 | 1,202.5 | 841 | 4,078 | 2012 |
| W7 | 2017–2023 | 21 | 33,430 | 323 | 1,203.0 | 447 | 4,018 | 2023 |
| W3 (1994–99) is not used. ‘Under 18’: respondents aged 15–17 (a few countries sample from 15 or 16); Part II uses adults only. |
Countries and regions
Region scheme
One region scheme covers every wave:
- India is its own group.
- Western Europe & Offshoots: Western European countries plus Australia, Canada, New Zealand and the USA. Cyprus and Greece are included.
- Middle East & North Africa: the Arab states, Iran and Turkey.
- Post-communist Central and Eastern Europe (Hungary, Poland and so on) is not in Western Europe. It sits in a separate region that is outside this comparison.
reg_n <- wvs |>
distinct(wave, macro_region, country_code) |>
count(macro_region, wave) |>
pivot_wider(names_from = wave, values_from = n, values_fill = 0)
presence <- wvs |>
distinct(country_code, macro_region, wave) |>
group_by(country_code, macro_region) |>
summarise(all4 = all(c("W4", "W5", "W6", "W7") %in% wave),
w567 = all(c("W5", "W6", "W7") %in% wave), .groups = "drop") |>
group_by(macro_region) |>
summarise(`In all 4 waves` = sum(all4), `In W5, W6 and W7` = sum(w567), `Ever surveyed` = n())
tab <- reg_n |>
left_join(presence, by = "macro_region") |>
mutate(macro_region = as.character(macro_region))
tab <- bind_rows(tab, tab |> summarise(across(where(is.numeric), sum)) |> mutate(macro_region = "Total"))
tab |>
rename(Region = macro_region) |>
kable(caption = "Number of countries per region and wave") |>
styled() |>
row_spec(nrow(tab), bold = TRUE)| Region | W4 | W5 | W6 | W7 | In all 4 waves | In W5, W6 and W7 | Ever surveyed |
|---|---|---|---|---|---|---|---|
| India | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| Middle East & North Africa | 8 | 6 | 13 | 9 | 5 | 5 | 15 |
| Western Europe & Offshoots | 4 | 16 | 8 | 11 | 1 | 6 | 18 |
| Total | 13 | 23 | 22 | 21 | 7 | 12 | 34 |
How to read this: the countries in each region change from wave to wave, so a change in a region’s average can reflect which countries were surveyed, not only changing attitudes. The section on region averages below shows how much this matters. Two cases need care:
- The report starts with W4 (India 2001). The earlier wave W3 (India 1995) is dropped: its only MENA country was Turkey. W4 has 8 MENA countries (Algeria, Egypt, Iran, Iraq, Jordan, Morocco, Saudi Arabia, Turkey).
- Western Europe is thin in W4 (4 countries: Canada, Spain, Sweden, USA) because the rest of Europe was surveyed that round by the separate European Values Study. It is also thin in W6 (8 countries).
- Israel (surveyed in W4 only) is removed from the data altogether.
cl <- wvs |>
count(macro_region, country_name, wave) |>
mutate(n = format(n, big.mark = ",")) |>
pivot_wider(names_from = wave, values_from = n, values_fill = "—") |>
select(macro_region, Country = country_name, any_of(c("W4", "W5", "W6", "W7"))) |>
arrange(macro_region, Country)
cl |>
select(-macro_region) |>
kable(caption = "Every country in the comparison, with respondents per wave", align = c("l", rep("r", 4))) |>
styled() |>
pack_rows(index = table(cl$macro_region)) |>
scroll_box(height = "500px")| Country | W4 | W5 | W6 | W7 |
|---|---|---|---|---|
| India | ||||
| India | 2,002 | 2,001 | 4,078 | 1,692 |
| Middle East & North Africa | ||||
| Algeria | 1,282 | — | 1,200 | — |
| Egypt | 3,000 | 3,051 | 1,523 | 1,200 |
| Iran | 2,532 | 2,667 | — | 1,499 |
| Iraq | 2,325 | 2,701 | 1,200 | 1,200 |
| Jordan | 1,223 | 1,200 | 1,200 | 1,203 |
| Kuwait | — | — | 1,303 | — |
| Lebanon | — | — | 1,200 | 1,200 |
| Libya | — | — | 2,131 | 1,196 |
| Morocco | 1,251 | 1,200 | 1,200 | 1,200 |
| Palestine | — | — | 1,000 | — |
| Qatar | — | — | 1,060 | — |
| Saudi Arabia | 1,502 | — | — | — |
| Tunisia | — | — | 1,205 | 1,208 |
| Turkey | 3,401 | 1,346 | 1,605 | 2,415 |
| Yemen | — | — | 1,000 | — |
| Western Europe & Offshoots | ||||
| Andorra | — | 1,003 | — | 1,004 |
| Australia | — | 1,421 | 1,477 | 1,813 |
| Canada | 1,931 | 2,164 | — | 4,018 |
| Cyprus | — | 1,050 | 1,000 | 1,000 |
| Finland | — | 1,014 | — | — |
| France | — | 1,001 | — | — |
| Germany | — | 2,064 | 2,046 | 1,528 |
| Greece | — | — | — | 1,200 |
| Italy | — | 1,012 | — | — |
| Netherlands | — | 1,050 | 1,902 | 2,145 |
| New Zealand | — | 954 | 841 | 1,057 |
| Northern Ireland | — | — | — | 447 |
| Norway | — | 1,025 | — | — |
| Spain | 1,209 | 1,200 | 1,189 | — |
| Sweden | 1,015 | 1,003 | 1,206 | — |
| Switzerland | — | 1,241 | — | — |
| United Kingdom | — | 1,041 | — | 2,609 |
| United States | 1,200 | 1,249 | 2,232 | 2,596 |
How region averages are weighted
Within a country: every estimate uses that country’s
own survey weight (W4 V245, W5 V259, W6
V258, W7 W_WEIGHT). These weights correct each
national sample towards the national population, mainly on sex, age and
education.
Across countries (a region): the region value is the simple mean of its country estimates. Every country counts once, whatever its population or sample size:
- The USA counts the same as New Zealand in Western Europe & Offshoots.
- Egypt counts the same as Qatar in MENA.
This equals pooling respondents with the WVS equilibrated
weight (S018), which rescales every country to N =
1,000.
India vs a region: India is always shown on its own. The comparison is therefore India vs the average country in each region, not India vs the region’s population. That is why every comparison table also gives India’s rank among the countries in the comparison (India plus the Western European and MENA countries surveyed in that wave). The rank does not depend on averaging.
Why not weight by population? Population-weighting would let the USA dominate Western Europe & Offshoots, and Egypt, Iran and Turkey dominate MENA. The WVS samples national publics, so the unit of comparison is the country.
The table below checks how much region averages move when only the countries surveyed in W5, W6 and W7 are kept (a balanced panel). The example indicator is “a strong leader who does not bother with parliament and elections is good”.
country_est <- wvs |>
filter(is.na(age) | age >= 18) |>
group_by(wave, country_code, macro_region) |>
filter(any(!is.na(reg_strongleader_good))) |>
summarise(est = 100 * wmean(reg_strongleader_good, weight), .groups = "drop")
panel_ids <- wvs |>
distinct(country_code, wave) |>
group_by(country_code) |>
filter(all(c("W5", "W6", "W7") %in% wave)) |>
distinct(country_code) |>
pull()
bind_rows(
country_est |> mutate(sample = "All countries"),
country_est |> filter(country_code %in% panel_ids) |> mutate(sample = "Balanced panel (W5–W7)")
) |>
filter(wave %in% c("W5", "W6", "W7")) |>
group_by(macro_region, sample, wave) |>
summarise(cell = sprintf("%.1f <small>(%d)</small>", mean(est), n()), .groups = "drop") |>
pivot_wider(names_from = wave, values_from = cell, values_fill = "—") |>
rename(Region = macro_region, Sample = sample) |>
kable(escape = FALSE, caption = "Strong leader is good (%): all countries vs balanced panel") |>
styled() |>
collapse_rows(columns = 1, valign = "top") |>
footnote(general = "Cells: mean of country estimates (number of countries). Adults only.",
general_title = "")| Region | Sample | W5 | W6 | W7 |
|---|---|---|---|---|
| India | All countries | 63.9 (1) | 64.9 (1) | 69.1 (1) |
| Balanced panel (W5–W7) | 63.9 (1) | 64.9 (1) | 69.1 (1) | |
| Middle East & North Africa | All countries | 35.9 (6) | 44.4 (12) | 48.3 (9) |
| Balanced panel (W5–W7) | 28.2 (5) | 50.2 (5) | 53.7 (5) | |
| Western Europe & Offshoots | All countries | 24.2 (16) | 31.1 (8) | 28.2 (11) |
| Balanced panel (W5–W7) | 29.3 (6) | 29.1 (6) | 33.3 (6) | |
| Cells: mean of country estimates (number of countries). Adults only. |
Demographic profile by region
Each cell is the mean of country-level weighted values, so it describes the average country in the region, as in Part II. All respondents are included.
Profile by wave
demo_country <- wvs |>
group_by(wave, country_code, macro_region) |>
summarise(
`Female %` = 100 * wmean(sex == 2, weight),
`Mean age` = wmean(age, weight),
`18–34 %` = 100 * wmean(age_cat3 == "18-34", weight),
`55+ %` = 100 * wmean(age_cat3 == "55+", weight),
`Lower education %` = 100 * wmean(educ3 == "Lower", weight),
`Higher education %`= 100 * wmean(educ3 == "Higher", weight),
`Upper / upper-middle class %` = 100 * wmean(class3 == "Upper / upper middle", weight),
`Lower class %` = 100 * wmean(class3 == "Lower", weight),
.groups = "drop")
for (w in c("W4", "W5", "W6", "W7")) {
cat("\n\n#### ", w, " (India ", india_years[w], ")\n\n", sep = "")
tab <- demo_country |>
filter(wave == w) |>
group_by(macro_region) |>
summarise(Countries = n(), across(`Female %`:`Lower class %`, ~ mean(.x, na.rm = TRUE))) |>
rename(Region = macro_region)
print(tab |>
kable(digits = 1, caption = paste("Demographic profile by region,", w)) |>
styled() |>
row_spec(which(tab$Region == "India"), bold = TRUE, background = "#FFF3D6"))
}W4 (India 2001)
| Region | Countries | Female % | Mean age | 18–34 % | 55+ % | Lower education % | Higher education % | Upper / upper-middle class % | Lower class % |
|---|---|---|---|---|---|---|---|---|---|
| India | 1 | 43.2 | 40.2 | 41.5 | 18.1 | 52.2 | 23.4 | 20.0 | 19.8 |
| Middle East & North Africa | 8 | 49.6 | 35.8 | 52.4 | 11.8 | 40.8 | 20.2 | 25.1 | 14.7 |
| Western Europe & Offshoots | 4 | 50.7 | 44.4 | 33.2 | 28.9 | 23.4 | 31.0 | 30.6 | 4.9 |
W5 (India 2006)
| Region | Countries | Female % | Mean age | 18–34 % | 55+ % | Lower education % | Higher education % | Upper / upper-middle class % | Lower class % |
|---|---|---|---|---|---|---|---|---|---|
| India | 1 | 43.1 | 41.4 | 36.7 | 19.9 | 50.4 | 22.0 | 20.1 | 27.2 |
| Middle East & North Africa | 6 | 50.2 | 37.1 | 49.8 | 13.5 | 47.0 | 17.1 | 21.3 | 12.2 |
| Western Europe & Offshoots | 16 | 51.9 | 46.5 | 27.4 | 33.6 | 17.1 | 31.2 | 30.4 | 3.8 |
W6 (India 2012)
| Region | Countries | Female % | Mean age | 18–34 % | 55+ % | Lower education % | Higher education % | Upper / upper-middle class % | Lower class % |
|---|---|---|---|---|---|---|---|---|---|
| India | 1 | 43.8 | 41.2 | 38.1 | 19.0 | 46.1 | 13.3 | 18.8 | 14.2 |
| Middle East & North Africa | 13 | 48.8 | 37.5 | 48.4 | 14.6 | 35.9 | 26.2 | 26.8 | 11.3 |
| Western Europe & Offshoots | 8 | 51.8 | 47.9 | 27.4 | 36.8 | 17.3 | 34.5 | 28.2 | 4.1 |
W7 (India 2023)
| Region | Countries | Female % | Mean age | 18–34 % | 55+ % | Lower education % | Higher education % | Upper / upper-middle class % | Lower class % |
|---|---|---|---|---|---|---|---|---|---|
| India | 1 | 48.9 | 37.8 | 46.2 | 18.8 | 33.9 | 27.2 | 25.7 | 10.1 |
| Middle East & North Africa | 9 | 49.8 | 39.9 | 40.1 | 18.3 | 39.1 | 28.3 | 20.1 | 11.3 |
| Western Europe & Offshoots | 11 | 52.5 | 49.4 | 25.1 | 40.7 | 18.3 | 38.6 | 31.0 | 5.1 |
Religion by region
rel_country <- wvs |>
filter(!is.na(religion_major)) |>
group_by(wave, country_code, macro_region, religion_major) |>
summarise(w = sum(weight), .groups = "drop_last") |>
mutate(pct = 100 * w / sum(w)) |>
ungroup() |>
complete(nesting(wave, country_code, macro_region), religion_major, fill = list(pct = 0))
rel_country |>
filter(wave == "W7") |>
group_by(macro_region, religion_major) |>
summarise(pct = mean(pct), .groups = "drop") |>
pivot_wider(names_from = religion_major, values_from = pct) |>
rename(Region = macro_region) |>
kable(digits = 1, caption = "Religious denomination by region, W7 (% of respondents, mean of countries)") |>
styled() |>
scroll_box(width = "100%")| Region | None | Catholic | Protestant | Orthodox | Jewish | Muslim | Hindu | Buddhist | Other Christian | Other |
|---|---|---|---|---|---|---|---|---|---|---|
| India | 0.0 | 1.2 | 0 | 0.0 | 0.0 | 9.5 | 83.0 | 1.4 | 0.0 | 4.8 |
| Middle East & North Africa | 0.7 | 3.4 | 0 | 1.0 | 0.0 | 94.4 | 0.0 | 0.0 | 0.5 | 0.1 |
| Western Europe & Offshoots | 36.8 | 21.3 | 12 | 15.6 | 0.6 | 4.3 | 0.8 | 0.7 | 6.6 | 1.4 |
The India samples
How each sample was drawn
tribble(
~` `, ~`2001 (W4)`, ~`2006 (W5)`, ~`2012 (W6)`, ~`2023 (W7)`,
"Fieldwork", "2001", "Dec 2006 – Jan 2007", "2012", "5 Jun – 2 Jul 2023",
"Organisation", "Not documented in docs/", "Lokniti Network (PI Sandeep Shastri)", "Sandeep Shastri (Jain University CERRSE); W7 report credits Lokniti-CSDS", "Lokniti-CSDS",
"Design", "Not documented in docs/", "Multistage stratified random sample from electoral rolls: 40 Lok Sabha seats → 80 assembly seats → 160 polling stations → voters", "Four-stage stratified random sample from electoral rolls (Lok Sabha seats → assembly seats → polling stations → voters). Plan: 320 seats, 19,444 target", "8 states chosen purposively; 15 assembly seats (PPS); 60 polling stations; random-walk households; age and gender quotas",
"States", "18", "18 of 28 (97% of population)", "17", "8: Bihar, Delhi, Haryana, Maharashtra, Punjab, Telangana, UP, West Bengal",
"Achieved N", "2,002", "2,001", "4,078", "1,692 (1,371 aged 18+)",
"Age range", "18+", "18+", "18+", "16+ (321 aged 16–17 by design)",
"Languages", "—", "10", "—", "5",
"Comparable as a national trend point?", "Yes", "Yes", "Yes", "Partly: 8 states, quota, 60 clusters"
) |>
kable(caption = "India sample designs (from docs/ in each wave folder)") |>
styled() |>
column_spec(1, bold = TRUE) |>
footnote(general = "Sources: W5 study description; W6 India sample design proposal; W7 India technical report (Lokniti-CSDS, July 2023). No India methodology document for W4 is in the project folder.",
general_title = "")| 2001 (W4) | 2006 (W5) | 2012 (W6) | 2023 (W7) | |
|---|---|---|---|---|
| Fieldwork | 2001 | Dec 2006 – Jan 2007 | 2012 | 5 Jun – 2 Jul 2023 |
| Organisation | Not documented in docs/ | Lokniti Network (PI Sandeep Shastri) | Sandeep Shastri (Jain University CERRSE); W7 report credits Lokniti-CSDS | Lokniti-CSDS |
| Design | Not documented in docs/ | Multistage stratified random sample from electoral rolls: 40 Lok Sabha seats → 80 assembly seats → 160 polling stations → voters | Four-stage stratified random sample from electoral rolls (Lok Sabha seats → assembly seats → polling stations → voters). Plan: 320 seats, 19,444 target | 8 states chosen purposively; 15 assembly seats (PPS); 60 polling stations; random-walk households; age and gender quotas |
| States | 18 | 18 of 28 (97% of population) | 17 | 8: Bihar, Delhi, Haryana, Maharashtra, Punjab, Telangana, UP, West Bengal |
| Achieved N | 2,002 | 2,001 | 4,078 | 1,692 (1,371 aged 18+) |
| Age range | 18+ | 18+ | 18+ | 16+ (321 aged 16–17 by design) |
| Languages | — | 10 | — | 5 |
| Comparable as a national trend point? | Yes | Yes | Yes | Partly: 8 states, quota, 60 clusters |
| Sources: W5 study description; W6 India sample design proposal; W7 India technical report (Lokniti-CSDS, July 2023). No India methodology document for W4 is in the project folder. |
Where the interviews were done
wvs |>
filter(is_india) |>
count(in_state = tidyr::replace_na(in_state, "Not recorded"), wave) |>
mutate(n = as.character(n)) |>
pivot_wider(names_from = wave, values_from = n, values_fill = "—") |>
rename(State = in_state, `2001` = W4, `2006` = W5, `2012` = W6, `2023` = W7) |>
arrange(State) |>
kable(caption = "India: interviews per state", align = c("l", rep("r", 4))) |>
styled() |>
footnote(general = "Bihar, Delhi, Haryana, Maharashtra, Punjab, Uttar Pradesh and West Bengal are the only states in all four waves.",
general_title = "")| State | 2001 | 2006 | 2012 | 2023 |
|---|---|---|---|---|
| Andhra Pradesh | 152 | 143 | 307 | — |
| Assam | 61 | 52 | — | — |
| Bihar | 147 | 197 | 332 | 241 |
| Chhattisgarh | 46 | 44 | 99 | — |
| Delhi | 43 | 38 | 94 | 229 |
| Gujarat | 96 | 102 | 259 | — |
| Haryana | 50 | 44 | 122 | 115 |
| Jharkhand | 65 | 52 | 160 | — |
| Karnataka | 112 | 103 | 239 | — |
| Kerala | 71 | 59 | 192 | — |
| Madhya Pradesh | 115 | 104 | 256 | — |
| Maharashtra | 194 | 201 | 280 | 202 |
| Not recorded | — | 1 | — | — |
| Odisha | 68 | 75 | 355 | — |
| Punjab | 49 | 53 | 151 | 122 |
| Rajasthan | 102 | 112 | 278 | — |
| Tamil Nadu | 129 | 165 | — | — |
| Telangana | — | — | — | 210 |
| Unlabelled (W6 code 356019) | — | — | 64 | — |
| Uttar Pradesh | 328 | 301 | 596 | 360 |
| West Bengal | 174 | 155 | 294 | 213 |
| Bihar, Delhi, Haryana, Maharashtra, Punjab, Uttar Pradesh and West Bengal are the only states in all four waves. |
Who was interviewed: unweighted and weighted
Comparing the raw sample with the weighted one shows what the weight corrects.
india <- wvs |>
filter(is_india) |>
mutate(age_band = ifelse(!is.na(age) & age < 18, "16–17", as.character(age_cat3)))
profile_vars <- c(Sex = "sex_f", Age = "age_band", Education = "educ3", Religion = "in_religion",
Caste = "in_caste", `Subjective class` = "class3", `Town size` = "urban_size",
`Party would vote for` = "in_party_vote")
prof <- imap_dfr(profile_vars, function(v, lab) {
india |>
filter(!is.na(.data[[v]])) |>
group_by(wave, level = as.character(.data[[v]])) |>
summarise(n = n(), w = sum(weight), .groups = "drop_last") |>
mutate(Unweighted = 100 * n / sum(n), Weighted = 100 * w / sum(w)) |>
ungroup() |>
mutate(group = lab)
})
prof_wide <- prof |>
mutate(cell = sprintf("%.1f / %.1f", Unweighted, Weighted),
group = factor(group, names(profile_vars)),
year = india_years[wave]) |>
select(group, level, year, cell) |>
pivot_wider(names_from = year, values_from = cell, values_fill = "—") |>
arrange(group) |>
select(group, level, any_of(unname(india_years)))
prof_wide |>
select(-group) |>
rename(` ` = level) |>
kable(caption = "India sample profile: unweighted % / weighted %", align = c("l", rep("c", 4))) |>
styled() |>
pack_rows(index = table(prof_wide$group)) |>
footnote(general = "Percentages are among respondents with a valid answer on that variable. Caste was not asked in 2006.",
general_title = "")| 2001 | 2006 | 2012 | 2023 | |
|---|---|---|---|---|
| Sex | ||||
| Female | 43.2 / 43.2 | 43.1 / 43.1 | 43.8 / 43.8 | 43.4 / 48.9 |
| Male | 56.8 / 56.8 | 56.9 / 56.9 | 56.2 / 56.2 | 56.6 / 51.1 |
| Age | ||||
| 18-34 | 41.5 / 41.5 | 36.7 / 36.7 | 38.1 / 38.1 | 32.6 / 43.6 |
| 35-54 | 40.4 / 40.4 | 43.5 / 43.5 | 42.9 / 42.9 | 33.0 / 33.0 |
| 55+ | 18.1 / 18.1 | 19.9 / 19.9 | 19.0 / 19.0 | 15.5 / 17.7 |
| 16–17 | — | — | — | 19.0 / 5.7 |
| Education | ||||
| Higher | 23.4 / 23.4 | 22.0 / 22.0 | 13.3 / 13.3 | 23.1 / 27.2 |
| Lower | 52.2 / 52.2 | 50.4 / 50.4 | 46.1 / 46.1 | 32.2 / 33.9 |
| Middle | 24.4 / 24.4 | 27.6 / 27.6 | 40.6 / 40.6 | 44.6 / 38.9 |
| Religion | ||||
| Christian | 3.0 / 3.0 | 3.2 / 3.2 | 1.6 / 1.6 | 1.0 / 1.2 |
| Hindu | 73.3 / 73.3 | 77.8 / 77.8 | 81.9 / 81.9 | 82.0 / 83.0 |
| Muslim | 11.0 / 11.0 | 8.3 / 8.3 | 11.0 / 11.0 | 10.4 / 9.5 |
| None | 6.7 / 6.7 | 3.0 / 3.0 | 0.4 / 0.4 | — |
| Other (Sikh, Buddhist, Jain, …) | 6.1 / 6.1 | 7.6 / 7.6 | 5.1 / 5.1 | 6.6 / 6.2 |
| Caste | ||||
| General / other | 43.5 / 43.5 | — | 32.6 / 32.6 | 31.6 / 32.0 |
| OBC | 38.5 / 38.5 | — | 40.6 / 40.6 | 44.3 / 44.0 |
| SC | 12.9 / 12.9 | — | 20.1 / 20.1 | 19.1 / 19.2 |
| ST | 5.2 / 5.2 | — | 6.7 / 6.7 | 5.0 / 4.8 |
| Subjective class | ||||
| Lower | 19.8 / 19.8 | 27.2 / 27.2 | 14.2 / 14.2 | 10.1 / 10.1 |
| Lower middle / working | 60.2 / 60.2 | 52.7 / 52.7 | 67.0 / 67.0 | 63.9 / 64.1 |
| Upper / upper middle | 20.0 / 20.0 | 20.1 / 20.1 | 18.8 / 18.8 | 26.0 / 25.7 |
| Town size | ||||
| 100k+ | 11.2 / 11.2 | 7.9 / 7.9 | 3.1 / 3.1 | 5.9 / 5.7 |
| 20k-100k | 19.0 / 19.0 | 2.8 / 2.8 | 6.4 / 6.4 | 10.5 / 9.9 |
| Under 20k | 69.8 / 69.8 | 89.3 / 89.3 | 90.6 / 90.6 | 83.6 / 84.3 |
| Party would vote for | ||||
| BJP | 28.1 / 28.1 | 21.5 / 21.5 | 28.2 / 28.2 | 46.0 / 45.1 |
| INC | 33.1 / 33.1 | 32.4 / 32.4 | 28.0 / 28.0 | 21.4 / 21.9 |
| Left | 7.1 / 7.1 | 8.9 / 8.9 | 5.5 / 5.5 | 3.9 / 4.0 |
| No party / blank / other | 0.3 / 0.3 | 8.2 / 8.2 | 2.2 / 2.2 | 9.2 / 9.3 |
| Other party | 31.5 / 31.5 | 28.9 / 28.9 | 36.2 / 36.2 | 19.6 / 19.7 |
| Percentages are among respondents with a valid answer on that variable. Caste was not asked in 2006. | ||||
The India weights
india |>
group_by(wave) |>
summarise(N = n(),
Min = min(weight), Max = max(weight), Mean = mean(weight),
`CV of weights` = sd(weight) / mean(weight),
`Design effect (1 + CV²)` = 1 + (sd(weight) / mean(weight))^2,
`Effective N` = sum(weight)^2 / sum(weight^2),
.groups = "drop") |>
mutate(wave = india_years[wave]) |>
rename(`Survey year` = wave) |>
kable(digits = 2, caption = "India: survey weights", format.args = list(big.mark = ",")) |>
styled() |>
footnote(general = "Design effect here is from weighting only. Clustering (e.g. 60 polling stations in 2023) adds more, which the public data do not let us estimate.",
general_title = "")| Survey year | N | Min | Max | Mean | CV of weights | Design effect (1 + CV²) | Effective N |
|---|---|---|---|---|---|---|---|
| 2001 | 2,002 | 1.00 | 1.00 | 1 | 0.00 | 1.00 | 2,002.0 |
| 2006 | 2,001 | 1.00 | 1.00 | 1 | 0.00 | 1.00 | 2,001.0 |
| 2012 | 4,078 | 1.00 | 1.00 | 1 | 0.00 | 1.00 | 4,078.0 |
| 2023 | 1,692 | 0.19 | 1.94 | 1 | 0.44 | 1.19 | 1,419.8 |
| Design effect here is from weighting only. Clustering (e.g. 60 polling stations in 2023) adds more, which the public data do not let us estimate. |
The 2001, 2006 and 2012 India samples are unweighted. Every weight is 1, so “weighted” India estimates for those years are raw sample percentages. The profile table above shows what that means:
- Men are 55–57% of respondents.
- In 2006 and 2012 about 90% live in places under 20,000 people. The 2011 Census puts India’s urban share at about 31%.
- Only 13% have higher education in 2012, against 22–23% in 2001 and 2006.
Only 2023 is weighted: 16–17 year olds are brought down from 19% to 6% of the sample, and the sex split is balanced. For India trends this means:
- Changes between 2012 and 2023 partly reflect a change from unweighted to weighted samples.
- Group breakdowns (by sex, education or town size) are more trustworthy than India totals, because they do not depend on the sample’s overall mix.
- A useful robustness check is to reweight 2001–2012 to Census sex × education × urban margins. This needs Census tables that are not in the project yet.
How often people give no answer
Share of respondents with no substantive answer (don’t know, refused, missing). Only countries where the item was asked are counted. The regional figures are the median country.
nr_items <- c(reg_democracy = "Democratic system good/bad",
reg_strongleader = "Strong leader good/bad",
reg_army = "Army rule good/bad",
conf_parliament = "Confidence in parliament",
lr_scale = "Left–right self-placement",
econ_income = "Income equality scale",
dem_importance = "Importance of democracy",
income10 = "Income decile")
nr <- imap_dfr(nr_items, function(lab, v) {
wvs |>
group_by(wave, country_code, macro_region) |>
filter(any(!is.na(.data[[v]]))) |>
summarise(nr = 100 * wmean(is.na(.data[[v]]), weight), .groups = "drop") |>
group_by(wave, macro_region) |>
summarise(nr = median(nr), .groups = "drop") |>
mutate(item = lab)
})
nr_wide <- nr |>
mutate(macro_region = factor(macro_region, focus_regions),
year = paste0(wave, " / ", india_years[wave]),
cell = sprintf("%.0f", nr)) |>
select(item, macro_region, year, cell) |>
pivot_wider(names_from = year, values_from = cell, values_fill = "—") |>
arrange(match(item, nr_items), macro_region)
nr_wide |>
select(-item) |>
rename(` ` = macro_region) |>
kable(caption = "No answer (%): India and the median country in each region", align = c("l", rep("c", 4))) |>
styled() |>
pack_rows(index = table(factor(nr_wide$item, levels = unique(nr_wide$item))))| W4 / 2001 | W5 / 2006 | W6 / 2012 | W7 / 2023 | |
|---|---|---|---|---|
| Democratic system good/bad | ||||
| India | 27 | 24 | 12 | 8 |
| Western Europe & Offshoots | 6 | 3 | 3 | 4 |
| Middle East & North Africa | 13 | 8 | 7 | 3 |
| Strong leader good/bad | ||||
| India | 28 | 29 | 13 | 9 |
| Western Europe & Offshoots | 4 | 4 | 6 | 4 |
| Middle East & North Africa | 12 | 11 | 9 | 4 |
| Army rule good/bad | ||||
| India | 29 | 31 | 16 | 14 |
| Western Europe & Offshoots | 4 | 3 | 4 | 3 |
| Middle East & North Africa | 13 | 14 | 10 | 3 |
| Confidence in parliament | ||||
| India | 25 | 25 | 12 | 8 |
| Western Europe & Offshoots | 3 | 4 | 3 | 2 |
| Middle East & North Africa | 11 | 6 | 7 | 3 |
| Left–right self-placement | ||||
| India | 46 | 51 | 11 | 28 |
| Western Europe & Offshoots | 10 | 10 | 11 | 9 |
| Middle East & North Africa | 54 | 36 | 32 | 5 |
| Income equality scale | ||||
| India | 19 | 18 | 7 | 3 |
| Western Europe & Offshoots | 1 | 1 | 2 | 1 |
| Middle East & North Africa | 2 | 2 | 3 | 0 |
| Importance of democracy | ||||
| India | — | 20 | 1 | 5 |
| Western Europe & Offshoots | — | 1 | 2 | 2 |
| Middle East & North Africa | — | 2 | 3 | 1 |
| Income decile | ||||
| India | 1 | 0 | 1 | 1 |
| Western Europe & Offshoots | 9 | 9 | 5 | 4 |
| Middle East & North Africa | 10 | 1 | 1 | 1 |
Implication for Part II: India’s non-response on regime questions was around 24–31% in 2001 and 2006, well above the median Western European or MENA country. Percentages among those who answer therefore describe a more opinionated subset of Indians. That is why the analysis reports show the no-answer share next to every estimate.
Questions used in this report
Each question below is listed under the report heading where it is used. The table gives how the answer was recorded in the survey, how this report codes it, and the variable name in each wave’s data file (“—” = not asked in that wave).
codebook <- readr::read_csv("D:/Populism and Democrary/World value survey/data/clean/wvs_codebook.csv",
show_col_types = FALSE)
fmt_conf <- "4-point Likert: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all"
fmt_reg <- "4-point Likert: 1 very good, 2 fairly good, 3 fairly bad, 4 very bad"
fmt_ess <- "10-point scale: 1 not an essential characteristic ... 10 an essential characteristic"
fmt_agr4 <- "4-point Likert: 1 strongly agree, 2 agree, 3 disagree, 4 strongly disagree"
fmt_agr5 <- "5-point Likert: 1 agree strongly ... 3 neither ... 5 strongly disagree"
fmt_freq <- "4-point frequency: 1 often, 2 sometimes, 3 rarely, 4 never"
qtab <- tribble(
~heading, ~var, ~question, ~format, ~coding,
# 1 trust
"Trust in institutions (§1, §5)", "conf_civilservice", "Confidence in the civil service", fmt_conf, "1–2 = trusts. Institution that uses power",
"Trust in institutions (§1, §5)", "conf_police", "Confidence in the police", fmt_conf, "1–2 = trusts. Uses power",
"Trust in institutions (§1, §5)", "conf_army", "Confidence in the armed forces", fmt_conf, "1–2 = trusts. Uses power",
"Trust in institutions (§1, §5)", "conf_govt", "Confidence in the government", fmt_conf, "1–2 = trusts. Uses power",
"Trust in institutions (§1, §5)", "conf_parliament", "Confidence in parliament", fmt_conf, "1–2 = trusts. Institution that checks power",
"Trust in institutions (§1, §5)", "conf_parties", "Confidence in political parties", fmt_conf, "1–2 = trusts. Checks power",
"Trust in institutions (§1, §5)", "conf_press", "Confidence in the press", fmt_conf, "1–2 = trusts. Checks power",
"Trust in institutions (§1, §5)", "conf_courts", "Confidence in the courts / justice system", fmt_conf, "1–2 = trusts. Checks power (not asked in W4)",
# 2 governance and accountability
"Governance and accountability (§2, §5)", "pm_aims1", "Aims of the country (1st choice): economic growth / strong defence / more say at work and in communities / beautiful cities", "Ranked: 1st choice of 4", "Growth 1st or 2nd = delivery yes; more say 1st or 2nd = accountability yes",
"Governance and accountability (§2, §5)", "pm_aims2", "Aims of the country (2nd choice)", "Ranked: 2nd choice of 4", "As above",
"Governance and accountability (§2, §5)", "pm_goals1", "Most important (1st choice): maintaining order / more say in government / fighting rising prices / free speech", "Ranked: 1st choice of 4", "Prices 1st or 2nd = delivery yes; more say 1st or 2nd = accountability yes",
"Governance and accountability (§2, §5)", "pm_goals2", "Most important (2nd choice)", "Ranked: 2nd choice of 4", "As above",
"Governance and accountability (§2, §5)", "pm_society1", "Most important (1st choice): stable economy / more humane society / ideas over money / fight against crime", "Ranked: 1st choice of 4", "Stable economy or fight against crime 1st or 2nd = delivery yes",
"Governance and accountability (§2, §5)", "pm_society2", "Most important (2nd choice)", "Ranked: 2nd choice of 4", "As above",
"Governance and accountability (§2, §5)", "act_petition", "Signing a petition", "3-point: 1 have done, 2 might do, 3 would never do", "1–2 = accountability yes",
"Governance and accountability (§2, §5)", "act_demonstration", "Attending peaceful demonstrations", "3-point: 1 have done, 2 might do, 3 would never do", "1–2 = accountability yes; also shown by type (§3)",
"Governance and accountability (§2, §5)", "act_boycott", "Joining in boycotts", "3-point: 1 have done, 2 might do, 3 would never do", "1–2 = accountability yes; also shown by type (§3)",
# 3 conceptions
"Conceptions of democracy (§3)", "dess_elections", "Essential to democracy: people choose their leaders in free elections", fmt_ess, "Liberal question: 7–10 = yes. 1–6 = Illiberal",
"Conceptions of democracy (§3)", "dess_civilrights", "Essential to democracy: civil rights protect people from state oppression", fmt_ess, "Liberal question: 7–10 = yes",
"Conceptions of democracy (§3)", "dess_women", "Essential to democracy: women have the same rights as men", fmt_ess, "Liberal question: 7–10 = yes",
"Conceptions of democracy (§3)", "reg_democracy", "Having a democratic political system", fmt_reg, "Liberal question: 1–2 (good) = yes",
"Conceptions of democracy (§3)", "dem_importance", "Importance of living in a democratically governed country", "10-point: 1 not at all ... 10 absolutely important", "Liberal question: 7–10 = yes",
"Conceptions of democracy (§3)", "dess_army", "Essential to democracy: the army takes over when government is incompetent", fmt_ess, "Authoritarian question: 7–10 = yes",
"Conceptions of democracy (§3)", "dess_religious", "Essential to democracy: religious authorities interpret the laws", fmt_ess, "Authoritarian question: 7–10 = yes",
"Conceptions of democracy (§3)", "dess_obey", "Essential to democracy: people obey their rulers", fmt_ess, "Authoritarian question: 7–10 = yes",
"Conceptions of democracy (§3)", "reg_army", "Having the army rule", fmt_reg, "Authoritarian question: 1–2 (good) = yes",
"Conceptions of democracy (§3)", "reg_strongleader", "Having a strong leader who does not have to bother with parliament and elections", fmt_reg, "Authoritarian question: 1–2 (good) = yes",
"Conceptions of democracy (§3)", "reg_experts", "Having experts, not government, make decisions", fmt_reg, "Authoritarian question: 1–2 (good) = yes",
"Conceptions of democracy (§3)", "fut_authority", "In the near future: greater respect for authority", "3-point: 1 good thing, 2 don't mind, 3 bad thing", "Authoritarian question: 1 (good thing) = yes",
# 4 state support
"State support (§4)", "econ_govresp", "Government should take more responsibility to ensure everyone is provided for ... people should take more responsibility", "10-point scale between the two statements", "1–4 = yes",
"State support (§4)", "econ_income", "Incomes should be made more equal ... we need larger income differences as incentives", "10-point scale between the two statements", "1–4 = yes",
"State support (§4)", "dess_taxrich", "Essential to democracy: governments tax the rich and subsidise the poor", fmt_ess, "7–10 = yes",
"State support (§4)", "dess_unemp", "Essential to democracy: people receive state aid for unemployment", fmt_ess, "7–10 = yes",
"State support (§4)", "dess_incomeequal", "Essential to democracy: the state makes people's incomes equal", fmt_ess, "7–10 = yes. State supporter = yes to 3 or more of these 5",
# background
"Background: groups", "sex", "Respondent's sex", "Binary: 1 male, 2 female", "Male / female",
"Background: groups", "age", "Age in years", "Numeric (years)", "18–34 / 35–54 / 55+",
"Background: groups", "educ_level", "Highest education attained", "Categorical: 9 levels (W4–W6); ISCED 0–8 (W7)", "Lower / middle / higher (WVS standard cut)",
"Background: groups", "party_vote_code", "Party would vote for if election tomorrow", "Categorical: country party list", "India: BJP / INC / Left / other party",
"Background: groups", "religion_raw", "Religious denomination", "Categorical: denomination list", "India: Hindu / Muslim / Christian / other / none",
"Background: groups", "in_caste_code", "Caste / community (India)", "Categorical: Lokniti caste list (W4); SC/ST/OBC/other (W6, W7)", "SC / ST / OBC / general & other"
)
src <- codebook |> select(var, W4, W5, W6, W7)
src <- bind_rows(src, tibble(var = "in_caste_code", W4 = "V242", W5 = NA, W6 = "V254", W7 = "Q290"))
qt <- qtab |>
left_join(src, by = "var") |>
mutate(across(W4:W7, ~ ifelse(is.na(.x) | .x == "", "—", .x)),
heading = factor(heading, unique(qtab$heading))) |>
arrange(heading)
qt |>
select(Question = question, `How the answer was recorded` = format, `How this report codes it` = coding,
W4, W5, W6, W7) |>
kable(caption = "Questions used, response format and coding") |>
styled() |>
add_header_above(c(" " = 3, "Variable name in the data file" = 4)) |>
pack_rows(index = table(qt$heading)) |>
column_spec(1, width = "18em") |>
scroll_box(height = "600px")| Question | How the answer was recorded | How this report codes it | W4 | W5 | W6 | W7 |
|---|---|---|---|---|---|---|
| Trust in institutions (§1, §5) | ||||||
| Confidence in the civil service | 4-point Likert: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Institution that uses power | V156 | V141 | V118 | Q74 |
| Confidence in the police | 4-point Likert: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Uses power | V152 | V136 | V113 | Q69 |
| Confidence in the armed forces | 4-point Likert: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Uses power | V148 | V132 | V109 | Q65 |
| Confidence in the government | 4-point Likert: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Uses power | V153 | V138 | V115 | Q71 |
| Confidence in parliament | 4-point Likert: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Institution that checks power | V155 | V140 | V117 | Q73 |
| Confidence in political parties | 4-point Likert: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Checks power | V154 | V139 | V116 | Q72 |
| Confidence in the press | 4-point Likert: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Checks power | V149 | V133 | V110 | Q66 |
| Confidence in the courts / justice system | 4-point Likert: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Checks power (not asked in W4) | — | V137 | V114 | Q70 |
| Governance and accountability (§2, §5) | ||||||
| Aims of the country (1st choice): economic growth / strong defence / more say at work and in communities / beautiful cities | Ranked: 1st choice of 4 | Growth 1st or 2nd = delivery yes; more say 1st or 2nd = accountability yes | V120 | V69 | V60 | Q152 |
| Aims of the country (2nd choice) | Ranked: 2nd choice of 4 | As above | V121 | V70 | V61 | Q153 |
| Most important (1st choice): maintaining order / more say in government / fighting rising prices / free speech | Ranked: 1st choice of 4 | Prices 1st or 2nd = delivery yes; more say 1st or 2nd = accountability yes | V122 | V71 | V62 | Q154 |
| Most important (2nd choice) | Ranked: 2nd choice of 4 | As above | V123 | V72 | V63 | Q155 |
| Most important (1st choice): stable economy / more humane society / ideas over money / fight against crime | Ranked: 1st choice of 4 | Stable economy or fight against crime 1st or 2nd = delivery yes | V124 | V73 | V64 | Q156 |
| Most important (2nd choice) | Ranked: 2nd choice of 4 | As above | V125 | V74 | V65 | Q157 |
| Signing a petition | 3-point: 1 have done, 2 might do, 3 would never do | 1–2 = accountability yes | V134 | V96 | V85 | Q209 |
| Attending peaceful demonstrations | 3-point: 1 have done, 2 might do, 3 would never do | 1–2 = accountability yes; also shown by type (§3) | V136 | V98 | V87 | Q211 |
| Joining in boycotts | 3-point: 1 have done, 2 might do, 3 would never do | 1–2 = accountability yes; also shown by type (§3) | V135 | V97 | V86 | Q210 |
| Conceptions of democracy (§3) | ||||||
| Essential to democracy: people choose their leaders in free elections | 10-point scale: 1 not an essential characteristic … 10 an essential characteristic | Liberal question: 7–10 = yes. 1–6 = Illiberal | — | V154 | V133 | Q243 |
| Essential to democracy: civil rights protect people from state oppression | 10-point scale: 1 not an essential characteristic … 10 an essential characteristic | Liberal question: 7–10 = yes | — | V157 | V136 | Q246 |
| Essential to democracy: women have the same rights as men | 10-point scale: 1 not an essential characteristic … 10 an essential characteristic | Liberal question: 7–10 = yes | — | V161 | V139 | Q249 |
| Having a democratic political system | 4-point Likert: 1 very good, 2 fairly good, 3 fairly bad, 4 very bad | Liberal question: 1–2 (good) = yes | V167 | V151 | V130 | Q238 |
| Importance of living in a democratically governed country | 10-point: 1 not at all … 10 absolutely important | Liberal question: 7–10 = yes | — | V162 | V140 | Q250 |
| Essential to democracy: the army takes over when government is incompetent | 10-point scale: 1 not an essential characteristic … 10 an essential characteristic | Authoritarian question: 7–10 = yes | — | V156 | V135 | Q245 |
| Essential to democracy: religious authorities interpret the laws | 10-point scale: 1 not an essential characteristic … 10 an essential characteristic | Authoritarian question: 7–10 = yes | — | V153 | V132 | Q242 |
| Essential to democracy: people obey their rulers | 10-point scale: 1 not an essential characteristic … 10 an essential characteristic | Authoritarian question: 7–10 = yes | — | — | V138 | Q248 |
| Having the army rule | 4-point Likert: 1 very good, 2 fairly good, 3 fairly bad, 4 very bad | Authoritarian question: 1–2 (good) = yes | V166 | V150 | V129 | Q237 |
| Having a strong leader who does not have to bother with parliament and elections | 4-point Likert: 1 very good, 2 fairly good, 3 fairly bad, 4 very bad | Authoritarian question: 1–2 (good) = yes | V164 | V148 | V127 | Q235 |
| Having experts, not government, make decisions | 4-point Likert: 1 very good, 2 fairly good, 3 fairly bad, 4 very bad | Authoritarian question: 1–2 (good) = yes | V165 | V149 | V128 | Q236 |
| In the near future: greater respect for authority | 3-point: 1 good thing, 2 don’t mind, 3 bad thing | Authoritarian question: 1 (good thing) = yes | V130 | V78 | V69 | Q45 |
| State support (§4) | ||||||
| Government should take more responsibility to ensure everyone is provided for … people should take more responsibility | 10-point scale between the two statements | 1–4 = yes | V143 | V118 | V98 | Q108 |
| Incomes should be made more equal … we need larger income differences as incentives | 10-point scale between the two statements | 1–4 = yes | V141 | V116 | V96 | Q106 |
| Essential to democracy: governments tax the rich and subsidise the poor | 10-point scale: 1 not an essential characteristic … 10 an essential characteristic | 7–10 = yes | — | V152 | V131 | Q241 |
| Essential to democracy: people receive state aid for unemployment | 10-point scale: 1 not an essential characteristic … 10 an essential characteristic | 7–10 = yes | — | V155 | V134 | Q244 |
| Essential to democracy: the state makes people’s incomes equal | 10-point scale: 1 not an essential characteristic … 10 an essential characteristic | 7–10 = yes. State supporter = yes to 3 or more of these 5 | — | — | V137 | Q247 |
| Background: groups | ||||||
| Respondent’s sex | Binary: 1 male, 2 female | Male / female | V223 | V235 | V240 | Q260 |
| Age in years | Numeric (years) | 18–34 / 35–54 / 55+ | V225 | V237 | V242 | Q262 |
| Highest education attained | Categorical: 9 levels (W4–W6); ISCED 0–8 (W7) | Lower / middle / higher (WVS standard cut) | V226 | V238 | V248 | Q275 |
| Party would vote for if election tomorrow | Categorical: country party list | India: BJP / INC / Left / other party | V220 | V231 | V228 | Q223 |
| Religious denomination | Categorical: denomination list | India: Hindu / Muslim / Christian / other / none | V184G | V185 | V144G | Q289 |
| Caste / community (India) | Categorical: Lokniti caste list (W4); SC/ST/OBC/other (W6, W7) | SC / ST / OBC / general & other | V242 | — | V254 | Q290 |
How the questions build the §3 conception types
Each question is scored yes or no. The liberal score and the authoritarian score are the shares of each set answered yes; the democracy score is the liberal score minus the authoritarian score (−1 to +1). The types are bands of the democracy score.
tribble(
~Question, ~`Counted as yes`, ~Set,
"Free elections essential to democracy", "7–10", "Liberal",
"Civil rights essential to democracy", "7–10", "Liberal",
"Women's equal rights essential to democracy", "7–10", "Liberal",
"Having a democratic political system", "Good (1–2)", "Liberal",
"Importance of living in a democracy", "7–10", "Liberal",
"Army takeover essential to democracy", "7–10", "Authoritarian",
"Religious authorities interpret the laws, essential to democracy", "7–10", "Authoritarian",
"People obey their rulers, essential to democracy (from 2012)", "7–10", "Authoritarian",
"Having the army rule", "Good (1–2)", "Authoritarian",
"Having a strong leader who does not bother with parliament and elections", "Good (1–2)", "Authoritarian",
"Having experts, not government, make decisions", "Good (1–2)", "Authoritarian",
"Greater respect for authority", "Good thing (1)", "Authoritarian"
) |>
kable(caption = "Questions used to build the three conception types") |>
styled() |>
column_spec(1, width = "24em") |>
footnote(general = c("Liberal (maximalist) = democracy score 0.5 or more, with free elections 7–10.",
"Minimalist (electoral) = democracy score above 0 and below 0.5, with free elections 7–10.",
"Illiberal = democracy score 0 or less, or free elections 1–6.",
"Classified if free elections (asked from 2006) and at least three questions in each set were answered.",
"State support is a separate yes/no dimension built from five other questions (§4); it cuts across these types."),
general_title = "")| Question | Counted as yes | Set |
|---|---|---|
| Free elections essential to democracy | 7–10 | Liberal |
| Civil rights essential to democracy | 7–10 | Liberal |
| Women’s equal rights essential to democracy | 7–10 | Liberal |
| Having a democratic political system | Good (1–2) | Liberal |
| Importance of living in a democracy | 7–10 | Liberal |
| Army takeover essential to democracy | 7–10 | Authoritarian |
| Religious authorities interpret the laws, essential to democracy | 7–10 | Authoritarian |
| People obey their rulers, essential to democracy (from 2012) | 7–10 | Authoritarian |
| Having the army rule | Good (1–2) | Authoritarian |
| Having a strong leader who does not bother with parliament and elections | Good (1–2) | Authoritarian |
| Having experts, not government, make decisions | Good (1–2) | Authoritarian |
| Greater respect for authority | Good thing (1) | Authoritarian |
| Liberal (maximalist) = democracy score 0.5 or more, with free elections 7–10. | ||
| Minimalist (electoral) = democracy score above 0 and below 0.5, with free elections 7–10. | ||
| Illiberal = democracy score 0 or less, or free elections 1–6. | ||
| Classified if free elections (asked from 2006) and at least three questions in each set were answered. | ||
| State support is a separate yes/no dimension built from five other questions (§4); it cuts across these types. |
Part II. Governance and accountability
1. How much do people trust the state, and the institutions that check it?
The question. Some institutions exercise state power: the civil service, police, army and government. Others check it: parliament, parties, the press and courts. How much do people trust each one, and do they trust the state more than the institutions that check it?
What is calculated.
- Each institution separately: % with a great deal or quite a lot of confidence.
- Trust gap. For each respondent: share of the 4 state institutions trusted minus share of the 4 checking institutions trusted, in percentage points. Positive means more trust in the state than in the institutions that check it.
q_table(tribble(
~var, ~question, ~format, ~mapped,
"conf_civilservice", "Confidence in the civil service", fmt_conf, "1–2 = trusts. State (uses power)",
"conf_police", "Confidence in the police", fmt_conf, "1–2 = trusts. State (uses power)",
"conf_army", "Confidence in the armed forces", fmt_conf, "1–2 = trusts. State (uses power)",
"conf_govt", "Confidence in the government", fmt_conf, "1–2 = trusts. State (uses power)",
"conf_parliament", "Confidence in parliament", fmt_conf, "1–2 = trusts. Checks power",
"conf_parties", "Confidence in political parties", fmt_conf, "1–2 = trusts. Checks power",
"conf_press", "Confidence in the press", fmt_conf, "1–2 = trusts. Checks power",
"conf_courts", "Confidence in the courts", fmt_conf, "1–2 = trusts. Checks power"))| Question | Response format | How it is counted / what it maps to | Asked in India | W4 | W5 | W6 | W7 |
|---|---|---|---|---|---|---|---|
| Confidence in the civil service | 4-point: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. State (uses power) | 2001, 2006, 2012, 2023 | V156 | V141 | V118 | Q74 |
| Confidence in the police | 4-point: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. State (uses power) | 2001, 2006, 2012, 2023 | V152 | V136 | V113 | Q69 |
| Confidence in the armed forces | 4-point: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. State (uses power) | 2001, 2006, 2012, 2023 | V148 | V132 | V109 | Q65 |
| Confidence in the government | 4-point: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. State (uses power) | 2001, 2006, 2012, 2023 | V153 | V138 | V115 | Q71 |
| Confidence in parliament | 4-point: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Checks power | 2001, 2006, 2012, 2023 | V155 | V140 | V117 | Q73 |
| Confidence in political parties | 4-point: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Checks power | 2001, 2006, 2012, 2023 | V154 | V139 | V116 | Q72 |
| Confidence in the press | 4-point: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Checks power | 2001, 2006, 2012, 2023 | V149 | V133 | V110 | Q66 |
| Confidence in the courts | 4-point: 1 a great deal, 2 quite a lot, 3 not very much, 4 none at all | 1–2 = trusts. Checks power | 2006, 2012, 2023 | — | V137 | V114 | Q70 |
Each institution
inst8 <- c(conf_civilservice_hi = "Civil service", conf_police_hi = "Police",
conf_army_hi = "Armed forces", conf_govt_hi = "Government",
conf_parliament_hi = "Parliament", conf_parties_hi = "Political parties",
conf_press_hi = "Press", conf_courts_hi = "Courts")
simple_trend(india_trend(inst8), caption = "India: % with a great deal or quite a lot of confidence") |>
pack_rows("Institutions that use power (the state)", 1, 4) |>
pack_rows("Institutions that check power", 5, 8)| 2001 | 2006 | 2012 | 2023 | |
|---|---|---|---|---|
| Institutions that use power (the state) | ||||
| Civil service | 49 | 54 | 61 | 81 |
| Police | 38 | 64 | 51 | 67 |
| Armed forces | 92 | 83 | 87 | 87 |
| Government | 56 | 55 | 50 | 65 |
| Institutions that check power | ||||
| Parliament | 55 | 62 | 58 | 74 |
| Political parties | 34 | 46 | 37 | 43 |
| Press | 70 | 76 | 72 | 66 |
| Courts | — | 69 | 64 | 74 |
trust8 <- imap_dfr(inst8, function(lab, v) {
country_est(v) |>
group_by(wave, macro_region) |>
summarise(est = mean(est), k = n(), .groups = "drop") |>
mutate(inst = lab)
}) |>
mutate(inst = factor(inst, inst8),
macro_region = factor(macro_region, focus_regions),
wave_short = factor(sub(" .*", "", wave_labels[wave]), c("W4", "W5", "W6", "W7")))
p <- ggplot(trust8, aes(wave_short, est, colour = macro_region, group = macro_region)) +
geom_line(linewidth = 0.8) +
geom_point_interactive(aes(fill = macro_region, data_id = paste(inst, macro_region, wave),
tooltip = sprintf("%s<br>%s, %s: %.0f%%<br>%d %s", inst, macro_region,
wave_labels[wave], est, k, ifelse(k == 1, "country", "countries"))),
shape = 21, colour = "white", size = 2.6, stroke = 0.7) +
facet_wrap(~ inst, ncol = 4) +
scale_colour_manual(values = region_cols, breaks = focus_regions) +
scale_fill_manual(values = region_cols, guide = "none") +
scale_y_continuous(limits = c(0, 100), breaks = c(0, 25, 50, 75, 100)) +
labs(title = "Confidence in each institution: India vs the two regions",
x = NULL, y = "% a great deal or quite a lot",
caption = paste0("Top row: institutions that use power (state). Bottom row: institutions that check it.
",
"Courts not asked in W4. Regions: mean of countries. India: W4 2001, W5 2006, W6 2012, W7 2023.")) +
theme_wvs(10) + theme(panel.spacing = unit(1.2, "lines"))
show_plot(p, 11, 6.5)The trust gap
simple_trend(india_trend(c(state_trust = "Trust in the state (% of 4 institutions trusted)",
check_trust = "Trust in checking institutions (% of 4 trusted)",
trust_gap = "Gap (points)")),
caption = "India: trust in the two sides of the state",
footnote_text = "2001: no question on the courts, so the checking figure uses 3 institutions.")| 2001 | 2006 | 2012 | 2023 | |
|---|---|---|---|---|
| Trust in the state (% of 4 institutions trusted) | 59 | 65 | 62 | 75 |
| Trust in checking institutions (% of 4 trusted) | 54 | 64 | 58 | 64 |
| Gap (points) | 5 | 1 | 4 | 11 |
| 2001: no question on the courts, so the checking figure uses 3 institutions. |
plot_india_lines(india_trend(c(state_trust = "Trust in the state", check_trust = "Trust in checking institutions")),
title = "India: trust in the state vs trust in the institutions that check it",
ylab = "% of institutions trusted",
caption = "Shaded bands: 95% CI. 2001 checking index excludes courts (not asked in W4), so 2001 is not comparable.")region_compare("trust_gap", caption = "Trust gap (state minus checking, percentage points), by region")| Region | W4 (1999–2004) | W5 (2004–09) | W6 (2010–16) | W7 (2017–23) |
|---|---|---|---|---|
| India | 4.7 (1) | 0.7 (1) | 4.2 (1) | 10.7 (1) |
| Western Europe & Offshoots | 21.6 (3) | 19.0 (16) | 18.2 (8) | 23.5 (11) |
| Middle East & North Africa | 22.0 (6) | 15.5 (4) | 15.1 (13) | 19.9 (8) |
| India’s rank among countries shown (1 = highest) | 9 of 10 | 21 of 21 | 21 of 22 | 20 of 20 |
| Cells: mean of country estimates (number of countries). W4 has no courts item, so its checking-trust index uses 3 institutions. |
plot_country_dots("trust_gap", title = "Trust gap by country: India is at or near the bottom in every wave",
xlab = "Trust in state minus trust in checking institutions (percentage points)")india_groups("trust_gap", caption = "India: trust gap (percentage points), by group")| 2001 | 2006 | 2012 | 2023 | |
|---|---|---|---|---|
| Sex | ||||
| Male | 3.5 (917) | 0.8 (962) | 4.3 (2145) | 10.1 (754) |
| Female | 7.0 (507) | 0.6 (571) | 4.1 (1506) | 11.4 (555) |
| Age | ||||
| 18-34 | 4.2 (622) | 0.9 (585) | 3.5 (1436) | 11.7 (531) |
| 35-54 | 5.4 (575) | 0.9 (662) | 4.9 (1546) | 9.3 (537) |
| 55+ | 4.6 (225) | -0.1 (287) | 4.2 (651) | 11.1 (241) |
| Education | ||||
| Lower | 4.5 (564) | 0.5 (687) | 4.9 (1551) | 9.6 (452) |
| Middle | 6.2 (410) | 0.2 (454) | 3.7 (1575) | 10.5 (487) |
| Higher | 3.6 (446) | 1.9 (390) | 3.9 (521) | 12.3 (368) |
| Party | ||||
| BJP | 4.1 (362) | 2.5 (353) | 5.0 (950) | 12.1 (513) |
| INC | 4.5 (349) | -1.1 (488) | 4.6 (904) | 11.2 (240) |
| Left | 4.7 (99) | -4.0 (132) | -2.7 (199) | n=44 |
| Other party | 1.8 (352) | 3.0 (453) | 3.8 (1216) | 10.2 (213) |
| Religion | ||||
| Hindu | 3.5 (1005) | 0.0 (1142) | 4.6 (2989) | 11.3 (1080) |
| Muslim | 6.1 (137) | -1.4 (125) | -0.5 (419) | 6.4 (130) |
| Christian | 11.7 (57) | n=42 | 0.3 (60) | n=13 |
| Other (Sikh, Buddhist, Jain, …) | -0.2 (105) | 5.4 (124) | 11.8 (169) | 10.2 (83) |
| Caste | ||||
| SC | -1.9 (155) | — | 4.8 (704) | 11.2 (253) |
| ST | 11.9 (66) | — | 1.5 (226) | 2.9 (64) |
| OBC | 4.1 (510) | — | 4.3 (1476) | 10.6 (562) |
| General / other | 5.6 (645) | — | 4.1 (1183) | 11.8 (430) |
| Cells: estimate (unweighted n). Cells with n < 50 are not estimated. | ||||
What this shows.
- India has the smallest trust gap of all countries shown in 2006, 2012 and 2023 (9th of 10 in 2001). In Western Europe and MENA people trust the state institutions (above all police and army) 15–24 points more than the institutions that check them. In India the two were level in 2006. The 2001 figure (4.7) is not comparable: W4 has no question on the courts, one of the most trusted institutions, so the checking index uses only parliament, parties and press. In the six common states the 2001 gap is 0.7.
- The gap is opening. It was 4 points in 2012 and 11 in 2023. It still opens when only the seven states surveyed every wave are used (9.6 points in 2023; see the robustness table in section 5). Trust in the state jumped in 2023 while trust in checking institutions did not.
- The gap is widespread within India. In 2023 it is 9–12 points in almost every group: sex, age, education, BJP and INC voters, all castes except ST. The shift is not driven by one constituency. Muslims have the smallest gap.
2. Governance, accountability, or both?
The question. The passage argued that Indians want both democracy and a strong, effective government. Do people want a state that delivers, a state that answers to them, or both? And are the two demands held by the same people?
What is calculated. Two yes/no measures for each respondent:
- Demand for governance (delivery), four questions: picks economic growth, fighting rising prices, a stable economy and the fight against crime among the top two aims (each counts as a separate yes). All four were asked in every survey from 2001.
- Demand for accountability, including willingness to act, five questions: picks more say in government decisions, and more say at work and in communities, among the top two aims; has signed or would sign a petition, has attended or would attend a peaceful demonstration, has joined or would join a boycott. All five were asked in every survey from 2001. Petition, demonstration and boycott record willingness to take part, so this side measures participation as well as preference. A robustness check below shows the groups without demonstrations and boycotts. (Free elections and civil rights as essentials of democracy belong to the conception types in section 3.)
A respondent demands governance (or accountability) if they say yes to at least half of the questions they answered (2 of 4 for delivery, 3 of 5 for accountability). Those who skipped more than one of the questions asked in their survey are not classified; that share is shown separately. Crossing the two demands gives four groups: Accountability-first (accountability only), Want both, Governance-first (governance only) and Neither.
How to read it. “Want both” is the group the passage describes. Governance-first is the group that cares about results more than about holding government to account. Three questions on each side come from “pick your top two” lists, so choosing more delivery aims leaves less room for accountability aims; the groups partly reflect that trade-off.
fmt_rank <- function(list) paste0("Ranked: 1st and 2nd choice from 4 options (", list, ")")
aims_list <- "economic growth / strong defence / more say at work and in communities / beautiful cities"
goals_list <- "maintaining order / more say in government / fighting rising prices / free speech"
society_list <- "stable economy / more humane society / ideas over money / fight against crime"
fmt_act <- "3-point: 1 have done, 2 might do, 3 would never do"
gov_side <- "Governance (delivery): yes on at least half of 4"
acc_side <- "Accountability, including willingness to act: yes on at least half of 5"
q_table(tribble(
~side, ~var, ~question, ~format, ~mapped,
gov_side, "pm_aims1+pm_aims2", "Aims of the country: a high level of economic growth", fmt_rank(aims_list), "Picked 1st or 2nd = yes",
gov_side, "pm_goals1+pm_goals2", "Most important for the country: fighting rising prices", fmt_rank(goals_list), "Picked 1st or 2nd = yes",
gov_side, "pm_society1+pm_society2", "Most important: a stable economy", fmt_rank(society_list), "Picked 1st or 2nd = yes",
gov_side, "pm_society1+pm_society2", "Most important: the fight against crime", fmt_rank(society_list), "Picked 1st or 2nd = yes",
acc_side, "pm_goals1+pm_goals2", "Most important for the country: giving people more say in important government decisions", fmt_rank(goals_list), "Picked 1st or 2nd = yes",
acc_side, "pm_aims1+pm_aims2", "Aims of the country: people have more say about how things are done at their jobs and in their communities", fmt_rank(aims_list), "Picked 1st or 2nd = yes",
acc_side, "act_petition", "Signing a petition", fmt_act, "1–2 (done or might do) = yes",
acc_side, "act_demonstration", "Attending peaceful demonstrations", fmt_act, "1–2 (done or might do) = yes",
acc_side, "act_boycott", "Joining in boycotts", fmt_act, "1–2 (done or might do) = yes"),
caption = "Questions used in this section")| Question | Response format | How it is counted / what it maps to | Asked in India | W4 | W5 | W6 | W7 |
|---|---|---|---|---|---|---|---|
| Governance (delivery): yes on at least half of 4 | |||||||
| Aims of the country: a high level of economic growth | Ranked: 1st and 2nd choice from 4 options (economic growth / strong defence / more say at work and in communities / beautiful cities) | Picked 1st or 2nd = yes | 2001, 2006, 2012, 2023 | V120 & V121 | V69 & V70 | V60 & V61 | Q152 & Q153 |
| Most important for the country: fighting rising prices | Ranked: 1st and 2nd choice from 4 options (maintaining order / more say in government / fighting rising prices / free speech) | Picked 1st or 2nd = yes | 2001, 2006, 2012, 2023 | V122 & V123 | V71 & V72 | V62 & V63 | Q154 & Q155 |
| Most important: a stable economy | Ranked: 1st and 2nd choice from 4 options (stable economy / more humane society / ideas over money / fight against crime) | Picked 1st or 2nd = yes | 2001, 2006, 2012, 2023 | V124 & V125 | V73 & V74 | V64 & V65 | Q156 & Q157 |
| Most important: the fight against crime | Ranked: 1st and 2nd choice from 4 options (stable economy / more humane society / ideas over money / fight against crime) | Picked 1st or 2nd = yes | 2001, 2006, 2012, 2023 | V124 & V125 | V73 & V74 | V64 & V65 | Q156 & Q157 |
| Accountability, including willingness to act: yes on at least half of 5 | |||||||
| Most important for the country: giving people more say in important government decisions | Ranked: 1st and 2nd choice from 4 options (maintaining order / more say in government / fighting rising prices / free speech) | Picked 1st or 2nd = yes | 2001, 2006, 2012, 2023 | V122 & V123 | V71 & V72 | V62 & V63 | Q154 & Q155 |
| Aims of the country: people have more say about how things are done at their jobs and in their communities | Ranked: 1st and 2nd choice from 4 options (economic growth / strong defence / more say at work and in communities / beautiful cities) | Picked 1st or 2nd = yes | 2001, 2006, 2012, 2023 | V120 & V121 | V69 & V70 | V60 & V61 | Q152 & Q153 |
| Signing a petition | 3-point: 1 have done, 2 might do, 3 would never do | 1–2 (done or might do) = yes | 2001, 2006, 2012, 2023 | V134 | V96 | V85 | Q209 |
| Attending peaceful demonstrations | 3-point: 1 have done, 2 might do, 3 would never do | 1–2 (done or might do) = yes | 2001, 2006, 2012, 2023 | V136 | V98 | V87 | Q211 |
| Joining in boycotts | 3-point: 1 have done, 2 might do, 3 would never do | 1–2 (done or might do) = yes | 2001, 2006, 2012, 2023 | V135 | V97 | V86 | Q210 |
gt_share <- function(d) {
d |>
group_by(wave) |>
summarise(across(c(g_accfirst, g_both, g_govfirst),
~ 100 * weighted.mean(.x, weight, na.rm = TRUE)),
g_neither = 100 - g_accfirst - g_both - g_govfirst,
unclassified = 100 * sum(weight[is.na(gtype)]) / sum(weight),
.groups = "drop")
}
gt_labels <- c(g_accfirst = "Accountability-first", g_both = "Want both",
g_govfirst = "Governance-first", g_neither = "Neither",
unclassified = "Could not be classified (% of all adults)")
india |>
filter(wave %in% c("W4", "W5", "W6", "W7")) |>
gt_share() |>
pivot_longer(-wave) |>
mutate(name = factor(gt_labels[name], gt_labels), year = india_years[wave]) |>
select(Group = name, year, value) |>
pivot_wider(names_from = year, values_from = value) |>
kable(digits = 1, caption = "India: governance and accountability groups (% of those classified)") |>
styled() |>
row_spec(5, italic = TRUE)| Group | 2001 | 2006 | 2012 | 2023 |
|---|---|---|---|---|
| Accountability-first | 4.8 | 6.1 | 6.0 | 14.3 |
| Want both | 43.0 | 43.7 | 35.0 | 32.7 |
| Governance-first | 47.3 | 44.7 | 53.4 | 44.0 |
| Neither | 4.9 | 5.5 | 5.6 | 9.0 |
| Could not be classified (% of all adults) | 25.8 | 29.2 | 17.3 | 13.9 |
gt_country <- wvs |>
filter(wave %in% c("W4", "W5", "W6", "W7")) |>
group_by(wave, country_code, macro_region) |>
filter(sum(!is.na(gtype)) > 100) |>
summarise(across(c(g_accfirst, g_both, g_govfirst),
~ 100 * weighted.mean(.x, weight, na.rm = TRUE)), .groups = "drop") |>
mutate(g_neither = 100 - g_accfirst - g_both - g_govfirst)
gt_country |>
group_by(wave, macro_region) |>
summarise(across(g_accfirst:g_neither, mean), k = n(), .groups = "drop") |>
pivot_longer(g_accfirst:g_neither) |>
mutate(cell = sprintf("%.1f", value), Group = factor(gt_labels[name], gt_labels),
col = paste(wave_labels[wave])) |>
select(Group, macro_region, col, cell) |>
pivot_wider(names_from = col, values_from = cell, values_fill = "—") |>
arrange(Group, factor(macro_region, focus_regions)) |>
select(-Group) |>
rename(Region = macro_region) |>
kable(caption = "Governance and accountability groups by region (%, mean of countries)") |>
styled() |>
pack_rows(index = table(factor(rep(gt_labels[1:4], each = 3), gt_labels[1:4])))| Region | W4 (1999–2004) | W5 (2004–09) | W6 (2010–16) | W7 (2017–23) |
|---|---|---|---|---|
| Accountability-first | ||||
| India | 4.8 | 6.1 | 6.0 | 14.3 |
| Western Europe & Offshoots | 17.2 | 17.3 | 11.0 | 16.7 |
| Middle East & North Africa | 9.7 | 4.6 | 5.0 | 6.7 |
| Want both | ||||
| India | 43.0 | 43.7 | 35.0 | 32.7 |
| Western Europe & Offshoots | 62.2 | 58.0 | 60.1 | 57.3 |
| Middle East & North Africa | 32.5 | 29.3 | 24.3 | 23.6 |
| Governance-first | ||||
| India | 47.3 | 44.7 | 53.4 | 44.0 |
| Western Europe & Offshoots | 18.9 | 22.1 | 27.1 | 23.9 |
| Middle East & North Africa | 52.6 | 62.1 | 65.0 | 61.7 |
| Neither | ||||
| India | 4.9 | 5.5 | 5.6 | 9.0 |
| Western Europe & Offshoots | 1.7 | 2.6 | 1.7 | 2.1 |
| Middle East & North Africa | 5.2 | 4.1 | 5.6 | 8.0 |
gt_cols <- c(`Accountability-first` = "#256abf", `Want both` = "#9ec5f4",
`Governance-first` = "#e34948", Neither = "#bdbcb6")
gt_plot <- bind_rows(
india |> filter(wave %in% c("W4", "W5", "W6", "W7")) |> gt_share() |> select(-unclassified) |>
mutate(macro_region = "India"),
gt_country |> filter(macro_region != "India") |>
group_by(wave, macro_region) |> summarise(across(g_accfirst:g_neither, mean), .groups = "drop") |>
mutate(macro_region = as.character(macro_region))
) |>
pivot_longer(g_accfirst:g_neither, names_to = "type", values_to = "pct") |>
mutate(type = factor(gt_labels[type], gt_labels[1:4]),
macro_region = factor(macro_region, rev(focus_regions)),
wave_lab = factor(wave_labels[wave], wave_labels))
p <- ggplot(gt_plot, aes(pct, macro_region, fill = type)) +
geom_col(width = 0.62, colour = "white", linewidth = 0.6, position = position_stack(reverse = TRUE)) +
geom_text(aes(label = ifelse(pct >= 7, sprintf("%.0f", pct), ""), group = type,
colour = type %in% c("Accountability-first", "Governance-first")),
position = position_stack(vjust = 0.5, reverse = TRUE), size = 3, show.legend = FALSE) +
facet_wrap(~ wave_lab, ncol = 1) +
scale_fill_manual(values = gt_cols) +
scale_colour_manual(values = c(`TRUE` = "white", `FALSE` = ink), guide = "none") +
scale_x_continuous(expand = expansion(mult = c(0, 0.03)), labels = function(x) paste0(x, "%")) +
labs(title = "Governance and accountability groups", x = NULL, y = NULL,
caption = "% of those classified. Regions: mean of countries.") +
theme_wvs() + theme(panel.grid.major.y = element_blank())
p # static: interactive bars need a newer ggiraph than is available for R 4.3
india_groups("g_both", caption = "India: 'Want both' (% of those classified), by group")| 2001 | 2006 | 2012 | 2023 | |
|---|---|---|---|---|
| Sex | ||||
| Male | 51.3 (932) | 50.3 (892) | 41.7 (1966) | 37.5 (693) |
| Female | 29.1 (554) | 32.3 (524) | 25.8 (1405) | 27.1 (496) |
| Age | ||||
| 18-34 | 43.6 (640) | 44.1 (564) | 37.6 (1307) | 33.4 (485) |
| 35-54 | 43.5 (614) | 43.1 (610) | 33.8 (1451) | 34.0 (488) |
| 55+ | 40.0 (230) | 43.8 (242) | 32.7 (599) | 27.9 (216) |
| Education | ||||
| Lower | 37.0 (625) | 33.9 (578) | 24.2 (1449) | 25.1 (380) |
| Middle | 41.4 (413) | 44.8 (440) | 43.6 (1426) | 32.1 (456) |
| Higher | 53.0 (443) | 56.9 (394) | 42.1 (494) | 41.0 (351) |
| Party | ||||
| BJP | 41.4 (370) | 50.2 (311) | 41.2 (876) | 34.0 (471) |
| INC | 48.2 (394) | 42.4 (465) | 34.4 (864) | 32.6 (224) |
| Left | 35.8 (95) | 31.6 (152) | 26.8 (198) | n=42 |
| Other party | 46.0 (359) | 48.4 (401) | 32.4 (1155) | 30.9 (195) |
| Religion | ||||
| Hindu | 41.8 (1082) | 43.6 (1046) | 34.9 (2724) | 32.0 (973) |
| Muslim | 35.6 (132) | 31.2 (128) | 36.7 (392) | 24.3 (118) |
| Christian | 54.0 (50) | 52.9 (51) | 47.5 (61) | n=14 |
| Other (Sikh, Buddhist, Jain, …) | 55.3 (103) | 45.3 (117) | 28.9 (180) | 47.8 (81) |
| Caste | ||||
| SC | 52.5 (160) | — | 29.6 (673) | 31.1 (227) |
| ST | 50.0 (68) | — | 38.5 (226) | 20.5 (59) |
| OBC | 38.9 (553) | — | 33.6 (1338) | 34.2 (503) |
| General / other | 44.6 (655) | — | 38.8 (1081) | 33.3 (400) |
| Cells: estimate (unweighted n). Cells with n < 50 are not estimated. | ||||
india_groups("g_govfirst", caption = "India: 'Governance-first' (% of those classified), by group")| 2001 | 2006 | 2012 | 2023 | |
|---|---|---|---|---|
| Sex | ||||
| Male | 39.2 (932) | 38.5 (892) | 46.9 (1966) | 37.8 (693) |
| Female | 61.0 (554) | 55.5 (524) | 62.5 (1405) | 51.2 (496) |
| Age | ||||
| 18-34 | 47.3 (640) | 46.5 (564) | 51.6 (1307) | 43.5 (485) |
| 35-54 | 45.8 (614) | 44.6 (610) | 53.9 (1451) | 42.6 (488) |
| 55+ | 51.3 (230) | 41.3 (242) | 55.8 (599) | 48.3 (216) |
| Education | ||||
| Lower | 55.2 (625) | 52.6 (578) | 63.1 (1449) | 49.8 (380) |
| Middle | 47.0 (413) | 43.0 (440) | 45.6 (1426) | 44.6 (456) |
| Higher | 36.3 (443) | 35.0 (394) | 47.6 (494) | 37.6 (351) |
| Party | ||||
| BJP | 53.0 (370) | 36.0 (311) | 47.4 (876) | 43.1 (471) |
| INC | 42.1 (394) | 47.1 (465) | 54.5 (864) | 36.2 (224) |
| Left | 52.6 (95) | 58.6 (152) | 62.1 (198) | n=42 |
| Other party | 44.8 (359) | 39.7 (401) | 56.1 (1155) | 46.5 (195) |
| Religion | ||||
| Hindu | 49.3 (1082) | 46.0 (1046) | 53.7 (2724) | 44.7 (973) |
| Muslim | 54.5 (132) | 58.6 (128) | 48.7 (392) | 47.7 (118) |
| Christian | 36.0 (50) | 27.5 (51) | 42.6 (61) | n=14 |
| Other (Sikh, Buddhist, Jain, …) | 32.0 (103) | 37.6 (117) | 64.4 (180) | 34.5 (81) |
| Caste | ||||
| SC | 42.5 (160) | — | 58.4 (673) | 44.9 (227) |
| ST | 33.8 (68) | — | 50.4 (226) | 49.6 (59) |
| OBC | 52.3 (553) | — | 54.0 (1338) | 40.7 (503) |
| General / other | 44.4 (655) | — | 50.7 (1081) | 47.0 (400) |
| Cells: estimate (unweighted n). Cells with n < 50 are not estimated. | ||||
plot_dumbbell("g_govfirst", waves = c("W4", "W5", "W6", "W7"),
title = "India: 'Governance-first' by group, 2001 → 2023",
xlab = "% governance-first (of those classified)")Robustness: accountability without demonstrations and boycotts
Willingness to protest depends on opportunity as well as preference: safety, mobility, time and social norms all hold some groups back, women above all. This check rebuilds the accountability side from the other three questions only (more say in government, more say at work, petition) and compares the groups.
gt_lv <- c("Accountability-first", "Want both", "Governance-first", "Neither")
gt_dist <- function(d, v) {
d |>
filter(!is.na(.data[[v]])) |>
group_by(g = factor(as.character(.data[[v]]), gt_lv)) |>
summarise(p = sum(weight), .groups = "drop") |>
mutate(p = 100 * p / sum(p))
}
gt_versions <- function(v) {
ind <- india |>
group_by(wave) |>
group_modify(~ gt_dist(.x, v)) |>
ungroup() |>
transmute(g, col = paste("India", india_years[wave]), p)
reg <- wvs |>
filter(wave == "W7", !is_india) |>
group_by(macro_region, country_code) |>
group_modify(~ gt_dist(.x, v)) |>
ungroup() |>
complete(nesting(macro_region, country_code), g, fill = list(p = 0)) |>
group_by(macro_region, g) |>
summarise(p = mean(p), .groups = "drop") |>
transmute(g, col = ifelse(grepl("Western", macro_region), "W. Europe 2023", "MENA 2023"), p)
bind_rows(ind, reg)
}
rob <- bind_rows(gt_versions("gtype") |> mutate(version = "With demonstrations and boycotts (main measure)"),
gt_versions("gtype_pref") |> mutate(version = "Without demonstrations and boycotts"))
rob_cols <- c(paste("India", india_years), "W. Europe 2023", "MENA 2023")
rob_w <- rob |>
mutate(col = factor(col, rob_cols)) |>
arrange(version, g, col) |>
pivot_wider(names_from = col, values_from = p)
rob_w |>
select(-version) |>
rename(` ` = g) |>
kable(digits = 1, caption = "Governance and accountability groups, with and without demonstrations and boycotts (%)") |>
styled() |>
pack_rows(index = table(rob_w$version))| India 2001 | India 2006 | India 2012 | India 2023 | W. Europe 2023 | MENA 2023 | |
|---|---|---|---|---|---|---|
| With demonstrations and boycotts (main measure) | ||||||
| Accountability-first | 4.8 | 6.1 | 6.0 | 14.3 | 16.7 | 6.7 |
| Want both | 43.0 | 43.7 | 35.0 | 32.7 | 57.3 | 23.6 |
| Governance-first | 47.3 | 44.7 | 53.4 | 44.0 | 23.9 | 61.7 |
| Neither | 4.9 | 5.5 | 5.6 | 9.0 | 2.1 | 8.0 |
| Without demonstrations and boycotts | ||||||
| Accountability-first | 6.8 | 6.7 | 6.8 | 14.7 | 17.2 | 9.2 |
| Want both | 34.1 | 37.5 | 30.6 | 25.4 | 54.1 | 22.7 |
| Governance-first | 55.8 | 51.4 | 58.3 | 51.4 | 27.0 | 62.5 |
| Neither | 3.3 | 4.3 | 4.3 | 8.5 | 1.6 | 5.5 |
bind_rows(
india |> filter(wave == "W7", !is.na(sex_f)) |> group_by(sex_f) |> group_modify(~ gt_dist(.x, "gtype")) |>
mutate(version = "With demonstrations and boycotts"),
india |> filter(wave == "W7", !is.na(sex_f)) |> group_by(sex_f) |> group_modify(~ gt_dist(.x, "gtype_pref")) |>
mutate(version = "Without")
) |>
ungroup() |>
mutate(col = paste0(sex_f, ": ", version)) |>
select(g, col, p) |>
pivot_wider(names_from = col, values_from = p) |>
arrange(g) |>
rename(` ` = g) |>
kable(digits = 1, caption = "India 2023: groups by sex, with and without demonstrations and boycotts (%)") |>
styled()| Male: With demonstrations and boycotts | Female: With demonstrations and boycotts | Male: Without | Female: Without | |
|---|---|---|---|---|
| Accountability-first | 16.0 | 12.4 | 16.4 | 12.7 |
| Want both | 37.5 | 27.1 | 26.2 | 24.4 |
| Governance-first | 37.8 | 51.2 | 49.1 | 54.0 |
| Neither | 8.7 | 9.3 | 8.2 | 8.9 |
What this shows.
- Governance-first is India’s largest group in every survey: 47% (2001), 45% (2006), 53% (2012), 44% (2023). India sits between Western Europe (24% in 2023) and MENA (62%).
- Accountability-first more than doubled in 2023, from about 6% in every earlier survey to 14%, close to Western Europe (17%) and well above MENA (7%).
- “Want both” has shrunk: 43–44% in 2001 and 2006, 35% in 2012, 33% in 2023. It remains far below Western Europe (57%) but above MENA (24%). The passage’s “want both” group exists, but it is not the majority and it is not growing.
- The four surveys are directly comparable: both sides use the same questions in every survey from 2001.
- Within India, 2023:
- Women are more often governance-first than men (51% vs 38%).
- Education matters: “want both” rises from 25% (less educated) to 41% (highly educated); governance-first falls from 50% to 38%.
- BJP voters are more often governance-first than INC voters (43% vs 36%).
- Muslims are less often “want both” than Hindus (24% vs 32%).
- Robustness: without demonstrations and boycotts the picture holds. Governance-first is higher everywhere (India 51% in 2023), India still sits between the two regions, and accountability-first still rises in 2023. The gender gap in governance-first narrows from 13 points (51% vs 38%) to 5 (54% vs 49%): most of it comes from women’s lower willingness or ability to protest, not from different preferences.
- Caution: 26% (2001) and 29% (2006) of Indian adults could not be classified (17% in 2012, 14% in 2023).
3. Liberal, minimalist or illiberal democrats?
The question. A minimalist conception of democracy asks for free elections and rejects authoritarian alternatives, but does not insist on civil rights, equal rights and a firm commitment to democracy. A maximalist liberal conception insists on all of these. An illiberal conception either does not treat free elections as essential, or accepts authoritarian ways of governing. Which do people hold?
What is calculated. Two sets of questions, each answered yes or no (see the table):
- Liberal questions (5): free elections, civil rights and women’s equal rights rated 7–10 as essential to democracy; a democratic political system is good; living in a democracy is important (7–10).
- Authoritarian questions (7): army takeover and religious authorities rated 7–10 as essential to democracy; “people obey their rulers” rated 7–10; army rule, a strong leader without parliament and elections, and experts deciding instead of government called good; greater respect for authority a good thing.
Each respondent gets two scores: the liberal score, the share of the liberal questions they answered yes, and the authoritarian score, the share of the authoritarian questions they answered yes. The democracy score is the liberal score minus the authoritarian score. It runs from −1 (yes to every authoritarian question, no to every liberal one) to +1 (the reverse). This follows Kirsch and Welzel’s measure of liberal minus authoritarian notions of democracy. The three types are bands of this one score:
- Liberal (maximalist): democracy score of 0.5 or more, with free elections rated 7–10. Liberal answers clearly outweigh authoritarian ones.
- Minimalist (electoral): democracy score above 0 but below 0.5, with free elections rated 7–10. They lean democratic, but only by a modest margin.
- Illiberal: democracy score of 0 or less (authoritarian answers match or outweigh liberal ones), or free elections rated 1–6.
A respondent is classified if they answered the free-elections question (asked from 2006) and at least three questions in each set. “People obey their rulers” was asked only from 2012, so in 2006 the authoritarian score uses six questions.
Also shown for each type: willingness to attend a peaceful demonstration or join a boycott, to see whether liberals act on their views. State support is a separate dimension (section 4).
lib_side <- "Liberal questions: liberal score = share answered yes"
aut_side <- "Authoritarian questions: authoritarian score = share answered yes"
show_side <- "Shown for each type (not used in the rules)"
q_table(tribble(
~side, ~var, ~question, ~format, ~mapped,
lib_side, "dess_elections", "Essential to democracy: people choose their leaders in free elections", fmt_ess, "7–10 = yes. 1–6 = Illiberal whatever the score",
lib_side, "dess_civilrights", "Essential to democracy: civil rights protect people from state oppression", fmt_ess, "7–10 = yes",
lib_side, "dess_women", "Essential to democracy: women have the same rights as men", fmt_ess, "7–10 = yes",
lib_side, "reg_democracy", "Having a democratic political system", fmt_reg, "1–2 (good) = yes",
lib_side, "dem_importance", "How important is it for you to live in a country that is governed democratically?", "10-point: 1 not at all important ... 10 absolutely important", "7–10 = yes",
aut_side, "dess_army", "Essential to democracy: the army takes over when government is incompetent", fmt_ess, "7–10 = yes",
aut_side, "dess_religious", "Essential to democracy: religious authorities ultimately interpret the laws", fmt_ess, "7–10 = yes",
aut_side, "dess_obey", "Essential to democracy: people obey their rulers", fmt_ess, "7–10 = yes (from 2012)",
aut_side, "reg_army", "Having the army rule", fmt_reg, "1–2 (good) = yes",
aut_side, "reg_strongleader", "Having a strong leader who does not have to bother with parliament and elections", fmt_reg, "1–2 (good) = yes",
aut_side, "reg_experts", "Having experts, not government, make decisions according to what they think is best for the country", fmt_reg, "1–2 (good) = yes",
aut_side, "fut_authority", "In the near future: greater respect for authority", "3-point: 1 good thing, 2 don't mind, 3 bad thing", "1 (good thing) = yes",
show_side, "act_demonstration", "Attending peaceful demonstrations", "3-point: 1 have done, 2 might do, 3 would never do", "1–2 = done or might do",
show_side, "act_boycott", "Joining in boycotts", "3-point: 1 have done, 2 might do, 3 would never do", "1–2 = done or might do"),
caption = "Questions used in this section")| Question | Response format | How it is counted / what it maps to | Asked in India | W4 | W5 | W6 | W7 |
|---|---|---|---|---|---|---|---|
| Liberal questions: liberal score = share answered yes | |||||||
| Essential to democracy: people choose their leaders in free elections | 10-point: 1 not an essential characteristic of democracy … 10 an essential characteristic | 7–10 = yes. 1–6 = Illiberal whatever the score | 2006, 2012, 2023 | — | V154 | V133 | Q243 |
| Essential to democracy: civil rights protect people from state oppression | 10-point: 1 not an essential characteristic of democracy … 10 an essential characteristic | 7–10 = yes | 2006, 2012, 2023 | — | V157 | V136 | Q246 |
| Essential to democracy: women have the same rights as men | 10-point: 1 not an essential characteristic of democracy … 10 an essential characteristic | 7–10 = yes | 2006, 2012, 2023 | — | V161 | V139 | Q249 |
| Having a democratic political system | 4-point: 1 very good, 2 fairly good, 3 fairly bad, 4 very bad | 1–2 (good) = yes | 2001, 2006, 2012, 2023 | V167 | V151 | V130 | Q238 |
| How important is it for you to live in a country that is governed democratically? | 10-point: 1 not at all important … 10 absolutely important | 7–10 = yes | 2006, 2012, 2023 | — | V162 | V140 | Q250 |
| Authoritarian questions: authoritarian score = share answered yes | |||||||
| Essential to democracy: the army takes over when government is incompetent | 10-point: 1 not an essential characteristic of democracy … 10 an essential characteristic | 7–10 = yes | 2006, 2012, 2023 | — | V156 | V135 | Q245 |
| Essential to democracy: religious authorities ultimately interpret the laws | 10-point: 1 not an essential characteristic of democracy … 10 an essential characteristic | 7–10 = yes | 2006, 2012, 2023 | — | V153 | V132 | Q242 |
| Essential to democracy: people obey their rulers | 10-point: 1 not an essential characteristic of democracy … 10 an essential characteristic | 7–10 = yes (from 2012) | 2012, 2023 | — | — | V138 | Q248 |
| Having the army rule | 4-point: 1 very good, 2 fairly good, 3 fairly bad, 4 very bad | 1–2 (good) = yes | 2001, 2006, 2012, 2023 | V166 | V150 | V129 | Q237 |
| Having a strong leader who does not have to bother with parliament and elections | 4-point: 1 very good, 2 fairly good, 3 fairly bad, 4 very bad | 1–2 (good) = yes | 2001, 2006, 2012, 2023 | V164 | V148 | V127 | Q235 |
| Having experts, not government, make decisions according to what they think is best for the country | 4-point: 1 very good, 2 fairly good, 3 fairly bad, 4 very bad | 1–2 (good) = yes | 2001, 2006, 2012, 2023 | V165 | V149 | V128 | Q236 |
| In the near future: greater respect for authority | 3-point: 1 good thing, 2 don’t mind, 3 bad thing | 1 (good thing) = yes | 2001, 2006, 2012, 2023 | V130 | V78 | V69 | Q45 |
| Shown for each type (not used in the rules) | |||||||
| Attending peaceful demonstrations | 3-point: 1 have done, 2 might do, 3 would never do | 1–2 = done or might do | 2001, 2006, 2012, 2023 | V136 | V98 | V87 | Q211 |
| Joining in boycotts | 3-point: 1 have done, 2 might do, 3 would never do | 1–2 = done or might do | 2001, 2006, 2012, 2023 | V135 | V97 | V86 | Q210 |
conc_vars <- c("con_liberal", "con_minimal", "con_illiberal")
conc_share <- function(d) {
d |>
group_by(wave) |>
summarise(across(all_of(conc_vars), ~ 100 * weighted.mean(.x, weight, na.rm = TRUE)),
unclassified = 100 * sum(weight[is.na(conception)]) / sum(weight),
.groups = "drop")
}
conc_labels <- c(con_liberal = "Liberal (maximalist)", con_minimal = "Minimalist (electoral)",
con_illiberal = "Illiberal", unclassified = "Could not be classified (% of all adults)")
india |>
filter(wave %in% c("W5", "W6", "W7")) |>
conc_share() |>
pivot_longer(-wave) |>
mutate(name = factor(conc_labels[name], conc_labels), year = india_years[wave]) |>
select(Type = name, year, value) |>
pivot_wider(names_from = year, values_from = value) |>
kable(digits = 1, caption = "India: conceptions of democracy (% of those classified)") |>
styled() |>
row_spec(4, italic = TRUE)| Type | 2006 | 2012 | 2023 |
|---|---|---|---|
| Liberal (maximalist) | 25.0 | 29.5 | 26.4 |
| Minimalist (electoral) | 35.8 | 34.7 | 40.7 |
| Illiberal | 39.2 | 35.9 | 32.9 |
| Could not be classified (% of all adults) | 25.9 | 1.8 | 9.5 |
conc_country <- wvs |>
filter(wave %in% c("W5", "W6", "W7")) |>
group_by(wave, country_code, macro_region) |>
filter(sum(!is.na(conception)) > 100) |>
summarise(across(all_of(conc_vars), ~ 100 * weighted.mean(.x, weight, na.rm = TRUE)), .groups = "drop")
conc_country |>
group_by(wave, macro_region) |>
summarise(across(all_of(conc_vars), mean), .groups = "drop") |>
pivot_longer(all_of(conc_vars)) |>
mutate(cell = sprintf("%.1f", value), Type = factor(conc_labels[name], conc_labels[1:3]),
col = wave_labels[wave]) |>
select(Type, macro_region, col, cell) |>
pivot_wider(names_from = col, values_from = cell, values_fill = "—") |>
arrange(Type, factor(macro_region, focus_regions)) |>
select(-Type) |>
rename(Region = macro_region) |>
kable(caption = "Conceptions of democracy by region (%, mean of countries)") |>
styled() |>
pack_rows(index = table(factor(rep(conc_labels[1:3], each = 3), conc_labels[1:3])))| Region | W5 (2004–09) | W6 (2010–16) | W7 (2017–23) |
|---|---|---|---|
| Liberal (maximalist) | |||
| India | 25.0 | 29.5 | 26.4 |
| Western Europe & Offshoots | 73.1 | 67.1 | 68.9 |
| Middle East & North Africa | 30.3 | 21.2 | 23.3 |
| Minimalist (electoral) | |||
| India | 35.8 | 34.7 | 40.7 |
| Western Europe & Offshoots | 13.3 | 16.6 | 15.7 |
| Middle East & North Africa | 42.9 | 38.2 | 36.7 |
| Illiberal | |||
| India | 39.2 | 35.9 | 32.9 |
| Western Europe & Offshoots | 13.6 | 16.3 | 15.3 |
| Middle East & North Africa | 26.8 | 40.5 | 40.0 |
conc_cols <- c(`Liberal (maximalist)` = "#0d366b", `Minimalist (electoral)` = "#256abf", Illiberal = "#86b6ef")
conc_plot <- bind_rows(
india |> filter(wave %in% c("W5", "W6", "W7")) |> conc_share() |> select(-unclassified) |>
mutate(macro_region = "India"),
conc_country |> filter(macro_region != "India") |>
group_by(wave, macro_region) |> summarise(across(all_of(conc_vars), mean), .groups = "drop") |>
mutate(macro_region = as.character(macro_region))
) |>
pivot_longer(all_of(conc_vars), names_to = "type", values_to = "pct") |>
mutate(type = factor(conc_labels[type], conc_labels[1:3]),
macro_region = factor(macro_region, rev(focus_regions)),
wave_lab = factor(wave_labels[wave], wave_labels))
p <- ggplot(conc_plot, aes(pct, macro_region, fill = type)) +
geom_col(width = 0.62, colour = "white", linewidth = 0.6, position = position_stack(reverse = TRUE)) +
geom_text(aes(label = ifelse(pct >= 6, sprintf("%.0f", pct), ""), group = type,
colour = type %in% conc_labels[1:2]),
position = position_stack(vjust = 0.5, reverse = TRUE), size = 3, show.legend = FALSE) +
facet_wrap(~ wave_lab, ncol = 1) +
scale_fill_manual(values = conc_cols) +
scale_colour_manual(values = c(`TRUE` = "white", `FALSE` = ink), guide = "none") +
scale_x_continuous(expand = expansion(mult = c(0, 0.03)), labels = function(x) paste0(x, "%")) +
labs(title = "Conceptions of democracy: liberal, minimalist, illiberal",
x = NULL, y = NULL,
caption = "% of those classified. Regions: mean of countries. Essentials of democracy were first asked in 2006.") +
theme_wvs() + theme(panel.grid.major.y = element_blank())
p # static: interactive bars need a newer ggiraph than is available for R 4.3
plot_country_dots("con_liberal", waves = c("W5", "W6", "W7"),
title = "Liberal (maximalist) conception, by country",
xlab = "% liberal (maximalist) (of those classified)")india_groups("con_liberal", caption = "India: liberal (maximalist) conception (% of those classified), by group")| 2006 | 2012 | 2023 | |
|---|---|---|---|
| Sex | |||
| Male | 25.7 (925) | 29.8 (2251) | 25.9 (718) |
| Female | 23.7 (557) | 29.0 (1750) | 27.0 (531) |
| Age | |||
| 18-34 | 25.4 (574) | 29.4 (1512) | 27.6 (512) |
| 35-54 | 25.6 (657) | 29.7 (1708) | 24.1 (510) |
| 55+ | 22.3 (251) | 29.1 (760) | 27.9 (227) |
| Education | |||
| Lower | 18.0 (616) | 25.2 (1842) | 21.7 (413) |
| Middle | 26.8 (456) | 32.8 (1616) | 25.5 (475) |
| Higher | 33.3 (408) | 34.0 (539) | 32.6 (359) |
| Party | |||
| BJP | 26.0 (331) | 22.5 (1035) | 25.6 (494) |
| INC | 26.6 (473) | 37.0 (1016) | 20.7 (231) |
| Left | 27.2 (158) | 46.0 (202) | n=39 |
| Other party | 22.2 (423) | 27.7 (1308) | 28.4 (210) |
| Religion | |||
| Hindu | 24.9 (1105) | 29.4 (3276) | 26.3 (1031) |
| Muslim | 29.1 (117) | 25.6 (441) | 28.6 (119) |
| Christian | 27.8 (54) | 52.3 (65) | n=14 |
| Other (Sikh, Buddhist, Jain, …) | 22.7 (110) | 30.7 (205) | 24.4 (82) |
| Caste | |||
| SC | — | 27.1 (771) | 29.8 (237) |
| ST | — | 40.5 (259) | 20.8 (65) |
| OBC | — | 29.0 (1616) | 24.3 (538) |
| General / other | — | 28.2 (1292) | 28.2 (409) |
| Cells: estimate (unweighted n). Cells with n < 50 are not estimated. | |||
india_groups("con_illiberal", caption = "India: illiberal conception (% of those classified), by group")| 2006 | 2012 | 2023 | |
|---|---|---|---|
| Sex | |||
| Male | 36.2 (925) | 35.0 (2251) | 35.2 (718) |
| Female | 44.2 (557) | 37.0 (1750) | 30.2 (531) |
| Age | |||
| 18-34 | 39.2 (574) | 36.1 (1512) | 31.5 (512) |
| 35-54 | 39.3 (657) | 35.5 (1708) | 34.7 (510) |
| 55+ | 39.0 (251) | 37.0 (760) | 32.9 (227) |
| Education | |||
| Lower | 48.1 (616) | 38.9 (1842) | 38.7 (413) |
| Middle | 32.5 (456) | 34.8 (1616) | 33.3 (475) |
| Higher | 33.3 (408) | 28.9 (539) | 26.1 (359) |
| Party | |||
| BJP | 36.6 (331) | 40.6 (1035) | 35.0 (494) |
| INC | 38.1 (473) | 31.5 (1016) | 40.9 (231) |
| Left | 24.1 (158) | 24.8 (202) | n=39 |
| Other party | 48.0 (423) | 35.9 (1308) | 30.7 (210) |
| Religion | |||
| Hindu | 39.1 (1105) | 36.4 (3276) | 32.2 (1031) |
| Muslim | 37.6 (117) | 39.0 (441) | 42.0 (119) |
| Christian | 44.4 (54) | 18.5 (65) | n=14 |
| Other (Sikh, Buddhist, Jain, …) | 33.6 (110) | 24.9 (205) | 33.2 (82) |
| Caste | |||
| SC | — | 34.4 (771) | 33.0 (237) |
| ST | — | 31.3 (259) | 29.6 (65) |
| OBC | — | 37.3 (1616) | 37.6 (538) |
| General / other | — | 36.8 (1292) | 27.0 (409) |
| Cells: estimate (unweighted n). Cells with n < 50 are not estimated. | |||
plot_dumbbell("con_illiberal", waves = c("W5", "W6", "W7"),
title = "India: illiberal share by group",
xlab = "% illiberal (of those classified)")Do liberals act on their views?
For each conception type: the % who have taken part, or would take part, in a peaceful demonstration or a boycott. India uses its own weight. The regions pool their countries with each country weighted equally (equilibrated weight).
cross_tab <- function(v, caption, rows = "conception") {
groups <- list(
`India 2006` = filter(india, wave == "W5"), `India 2012` = filter(india, wave == "W6"),
`India 2023` = filter(india, wave == "W7"),
`W. Europe 2023` = filter(wvs, wave == "W7", macro_region == "Western Europe & Offshoots"),
`MENA 2023` = filter(wvs, wave == "W7", macro_region == "Middle East & North Africa"))
imap_dfr(groups, function(d, nm) {
w <- if (grepl("India", nm)) "weight" else "weight_eq"
d |>
filter(!is.na(.data[[rows]])) |>
group_by(row = .data[[rows]]) |>
group_modify(~ wstat(.x[[v]], .x[[w]])) |>
ungroup() |>
mutate(sample = nm)
}) |>
mutate(cell = ifelse(n < 50, sprintf("<small>n=%d</small>", n), sprintf("%.1f <small>(%d)</small>", est * 100, n)),
sample = factor(sample, c("India 2006", "India 2012", "India 2023", "W. Europe 2023", "MENA 2023"))) |>
select(row, sample, cell) |>
rename(` ` = row) |>
pivot_wider(names_from = sample, values_from = cell, values_fill = "—") |>
kable(escape = FALSE, caption = caption, align = c("l", rep("c", 5))) |>
styled() |>
footnote(general = "Cells: % (unweighted n). Cells with n < 50 are not estimated.", general_title = "")
}
cross_tab("acc_demo", "Has attended or might attend a peaceful demonstration (%), by conception")| India 2006 | India 2012 | India 2023 | W. Europe 2023 | MENA 2023 | |
|---|---|---|---|---|---|
| Liberal (maximalist) | 53.6 (332) | 55.2 (1057) | 57.0 (321) | 75.8 (12663) | 44.1 (2805) |
| Minimalist (electoral) | 49.9 (465) | 44.8 (1188) | 53.2 (474) | 58.8 (2752) | 34.5 (4095) |
| Illiberal | 46.7 (480) | 43.7 (1160) | 63.7 (383) | 58.5 (2915) | 32.1 (4593) |
| Cells: % (unweighted n). Cells with n < 50 are not estimated. |
cross_tab("acc_boycott", "Has joined or might join a boycott (%), by conception")| India 2006 | India 2012 | India 2023 | W. Europe 2023 | MENA 2023 | |
|---|---|---|---|---|---|
| Liberal (maximalist) | 55.2 (337) | 47.7 (1032) | 40.6 (314) | 66.8 (12564) | 40.4 (2785) |
| Minimalist (electoral) | 46.0 (474) | 35.8 (1164) | 40.5 (463) | 49.0 (2742) | 31.5 (4079) |
| Illiberal | 45.9 (481) | 34.4 (1148) | 48.7 (366) | 50.4 (2912) | 29.9 (4579) |
| Cells: % (unweighted n). Cells with n < 50 are not estimated. |
What this shows.
- Minimalists are India’s largest group. In 2023, 41% of Indians are minimalist, 26% liberal and 33% illiberal. Western Europe is the mirror image: 69% liberal, 16% minimalist, 15% illiberal. MENA is closest to India (23% liberal, 37% minimalist, 40% illiberal).
- India is moving towards the minimalist position. Minimalists rose from 36% (2006, 2012) to 41% (2023), while illiberals fell from 39% to 33%. The liberal share has stayed at 25–30%.
- Within India, 2023:
- Education matters most: 33% of the highly educated are liberal, against 22% of the less educated; illiberals are 26% and 39%.
- Muslims are more often illiberal than Hindus (42% vs 32%).
- INC voters are now more often illiberal than BJP voters (41% vs 35%); in 2012 it was the other way round (32% vs 41%).
- Women are slightly less often illiberal than men (30% vs 35%); in 2006 they were more often illiberal (44% vs 36%).
- Do liberals act on their views? In Western Europe yes, in India no longer. In Western Europe 76% of liberals would attend a peaceful demonstration, against 59% of illiberals. In India liberals were slightly ahead in 2006 and 2012 (55% vs 44% in 2012), but in 2023 illiberals are more willing to demonstrate (64% vs 57%) and to join a boycott (49% vs 41%).
- Caution: 26% of Indian adults in 2006 and 10% in 2023 could not be classified (2% in 2012).
4. State support: a separate dimension
The question. Do people want the state to look after those who cannot look after themselves, by taking responsibility for people’s needs, narrowing income gaps, taxing the rich and helping the unemployed? This is kept apart from the conception types in section 3, so that a liberal, a minimalist or an illiberal can each be for or against state support.
What is calculated. Each of the five questions in the table below counts as a “yes” to state support.
A respondent is a state supporter if they say yes to at least 3 of the 5. In 2006 only four were asked, so it is 3 of 4. 2001 had only two, so it cannot be classified. Only people who answered every question asked in their survey are classified.
q_table(tribble(
~var, ~question, ~format, ~mapped,
"econ_govresp", "Government should take more responsibility to ensure that everyone is provided for (1) ... people should take more responsibility to provide for themselves (10)", "10-point scale between the two statements", "1–4 = yes to state support",
"econ_income", "Incomes should be made more equal (1) ... we need larger income differences as incentives (10)", "10-point scale between the two statements", "1–4 = yes to state support",
"dess_taxrich", "Essential to democracy: governments tax the rich and subsidise the poor", fmt_ess, "7–10 = yes to state support",
"dess_unemp", "Essential to democracy: people receive state aid for unemployment", fmt_ess, "7–10 = yes to state support",
"dess_incomeequal", "Essential to democracy: the state makes people's incomes equal", fmt_ess, "7–10 = yes to state support"),
caption = "Questions used in this section. State supporter = yes to at least 3")| Question | Response format | How it is counted / what it maps to | Asked in India | W4 | W5 | W6 | W7 |
|---|---|---|---|---|---|---|---|
| Government should take more responsibility to ensure that everyone is provided for (1) … people should take more responsibility to provide for themselves (10) | 10-point scale between the two statements | 1–4 = yes to state support | 2001, 2006, 2012, 2023 | V143 | V118 | V98 | Q108 |
| Incomes should be made more equal (1) … we need larger income differences as incentives (10) | 10-point scale between the two statements | 1–4 = yes to state support | 2001, 2006, 2012, 2023 | V141 | V116 | V96 | Q106 |
| Essential to democracy: governments tax the rich and subsidise the poor | 10-point: 1 not an essential characteristic of democracy … 10 an essential characteristic | 7–10 = yes to state support | 2006, 2012, 2023 | — | V152 | V131 | Q241 |
| Essential to democracy: people receive state aid for unemployment | 10-point: 1 not an essential characteristic of democracy … 10 an essential characteristic | 7–10 = yes to state support | 2006, 2012, 2023 | — | V155 | V134 | Q244 |
| Essential to democracy: the state makes people’s incomes equal | 10-point: 1 not an essential characteristic of democracy … 10 an essential characteristic | 7–10 = yes to state support | 2012, 2023 | — | — | V137 | Q247 |
ss_labels <- c(ss_govresp_yes = "Government should take more responsibility",
ss_income_yes = "Incomes should be made more equal",
ss_taxrich_yes = "Essential: tax the rich, subsidise the poor",
ss_unemp_yes = "Essential: state aid for the unemployed",
ss_incomeeq_yes = "Essential: state makes incomes equal",
state_support = "State supporter (yes to 3 or more)")
tr_ss <- india_trend(as.list(ss_labels))
show_trend(tr_ss, caption = "India: % saying yes to each state-support question [95% CI]")| Item | 2001 | 2006 | 2012 | 2023 |
|---|---|---|---|---|
| Government should take more responsibility |
51.5 [49.1, 53.9] no answer 18% |
48.4 [45.9, 50.8] no answer 18% |
66.7 [65.2, 68.2] no answer 9% |
39.5 [36.8, 42.2] no answer 2% |
| Incomes should be made more equal |
54.4 [52.0, 56.9] no answer 19% |
48.0 [45.6, 50.4] no answer 18% |
71.1 [69.7, 72.6] no answer 7% |
20.6 [18.4, 22.9] no answer 3% |
| Essential: tax the rich, subsidise the poor | — |
72.7 [70.5, 74.8] no answer 16% |
65.0 [63.5, 66.5] no answer 2% |
69.4 [66.8, 72.1] no answer 11% |
| Essential: state aid for the unemployed | — |
73.5 [71.4, 75.7] no answer 17% |
69.0 [67.6, 70.5] no answer 2% |
66.0 [63.3, 68.7] no answer 7% |
| Essential: state makes incomes equal | — | — |
64.3 [62.8, 65.8] no answer 2% |
54.7 [51.8, 57.6] no answer 12% |
| State supporter (yes to 3 or more) | — |
46.3 [43.6, 49.0] no answer 34% |
71.6 [70.1, 73.1] no answer 13% |
52.4 [49.4, 55.5] no answer 19% |
plot_india_lines(tr_ss |> filter(var %in% c("ss_govresp_yes", "ss_income_yes")),
title = "India: what the government should do (% yes)", ylab = "% yes")plot_india_lines(tr_ss |> filter(var %in% c("ss_taxrich_yes", "ss_unemp_yes", "ss_incomeeq_yes")),
title = "India: what democracy must guarantee (% yes)", ylab = "% yes")ss_pan <- imap_dfr(ss_labels, function(lab, v) {
country_est(v) |>
group_by(wave, macro_region) |>
summarise(est = mean(est), k = n(), .groups = "drop") |>
mutate(item = lab)
}) |>
mutate(item = factor(item, ss_labels),
macro_region = factor(macro_region, focus_regions),
wave_short = factor(sub(" .*", "", wave_labels[wave]), c("W4", "W5", "W6", "W7")))
p <- ggplot(ss_pan, aes(wave_short, est, colour = macro_region, group = macro_region)) +
geom_line(linewidth = 0.8) +
geom_point_interactive(aes(fill = macro_region, data_id = paste(item, macro_region, wave),
tooltip = sprintf("%s<br>%s, %s: %.0f%%<br>%d %s", item, macro_region,
wave_labels[wave], est, k, ifelse(k == 1, "country", "countries"))),
shape = 21, colour = "white", size = 2.6, stroke = 0.7) +
facet_wrap(~ item, ncol = 3, labeller = label_wrap_gen(32)) +
scale_colour_manual(values = region_cols, breaks = focus_regions) +
scale_fill_manual(values = region_cols, guide = "none") +
scale_y_continuous(limits = c(0, 100), breaks = c(0, 25, 50, 75, 100)) +
labs(title = "State support, question by question: India vs the two regions",
x = NULL, y = "% yes",
caption = paste0("Regions: mean of countries. India: W4 2001, W5 2006, W6 2012, W7 2023.
",
"Essentials questions start in W5; state makes incomes equal starts in W6.")) +
theme_wvs(10) + theme(panel.spacing = unit(1.2, "lines"))
show_plot(p, 11, 6.5)imap_dfr(ss_labels, function(lab, v) {
country_est(v) |>
group_by(wave) |>
mutate(r = rank(-est, ties.method = "min"), N = n()) |>
ungroup() |>
filter(country_code == 356) |>
transmute(item = lab, wave, cell = sprintf("%.0f%%<br><small>%d of %d</small>", est, r, N))
}) |>
arrange(wave) |>
mutate(col = wave_labels[wave]) |>
select(Question = item, col, cell) |>
pivot_wider(names_from = col, values_from = cell, values_fill = "—") |>
arrange(match(Question, ss_labels)) |>
(\(x) kable(x, escape = FALSE, align = c("l", rep("c", ncol(x) - 1)),
caption = "India on each state-support question: % yes and rank among all countries shown (1 = most supportive)"))() |>
styled() |>
footnote(general = "Countries: India plus every Western European and MENA country surveyed in that wave.",
general_title = "")| Question | W4 (1999–2004) | W5 (2004–09) | W6 (2010–16) | W7 (2017–23) |
|---|---|---|---|---|
| Government should take more responsibility |
52% 7 of 13 |
48% 9 of 23 |
67% 4 of 22 |
39% 16 of 21 |
| Incomes should be made more equal |
54% 2 of 12 |
48% 4 of 23 |
71% 1 of 22 |
21% 19 of 21 |
| Essential: tax the rich, subsidise the poor | — |
73% 3 of 21 |
65% 6 of 22 |
69% 5 of 21 |
| Essential: state aid for the unemployed | — |
74% 11 of 21 |
69% 11 of 22 |
66% 11 of 21 |
| Essential: state makes incomes equal | — | — |
64% 4 of 22 |
55% 7 of 21 |
| State supporter (yes to 3 or more) | — |
46% 9 of 21 |
72% 4 of 22 |
52% 12 of 21 |
| Countries: India plus every Western European and MENA country surveyed in that wave. |
region_compare("state_support", caption = "State supporters (%), India and regions")| Region | W5 (2004–09) | W6 (2010–16) | W7 (2017–23) |
|---|---|---|---|
| India | 46.3 (1) | 71.6 (1) | 52.4 (1) |
| Western Europe & Offshoots | 32.1 (14) | 42.3 (8) | 44.6 (11) |
| Middle East & North Africa | 53.6 (6) | 54.3 (13) | 58.2 (9) |
| India’s rank among countries shown (1 = highest) | 9 of 21 | 4 of 22 | 12 of 21 |
| Cells: mean of country estimates (number of countries). |
plot_region_lines("state_support", title = "State supporters: India, Western Europe and MENA",
ylab = "% state supporters",
caption = "Regions: mean of country estimates. Not classifiable in 2001 (only two questions asked).")plot_country_dots("state_support", waves = c("W5", "W6", "W7"),
title = "State supporters, by country", xlab = "% state supporters")india_groups("state_support", caption = "India: state supporters (%), by group")| 2006 | 2012 | 2023 | |
|---|---|---|---|
| Sex | |||
| Male | 44.7 (838) | 70.5 (2044) | 50.4 (651) |
| Female | 48.9 (489) | 73.2 (1520) | 54.8 (466) |
| Age | |||
| 18-34 | 46.1 (523) | 70.3 (1373) | 50.1 (457) |
| 35-54 | 46.3 (581) | 71.8 (1534) | 53.1 (457) |
| 55+ | 46.6 (223) | 73.7 (638) | 57.5 (203) |
| Education | |||
| Lower | 52.2 (529) | 74.3 (1610) | 55.3 (360) |
| Middle | 46.7 (411) | 69.8 (1457) | 49.1 (428) |
| Higher | 37.0 (381) | 68.2 (493) | 53.6 (327) |
| Party | |||
| BJP | 41.3 (303) | 65.4 (949) | 48.2 (434) |
| INC | 50.5 (412) | 79.1 (907) | 50.5 (207) |
| Left | 46.8 (141) | 76.3 (186) | n=36 |
| Other party | 44.3 (384) | 70.5 (1185) | 54.3 (190) |
| Religion | |||
| Hindu | 44.3 (995) | 70.5 (2902) | 52.9 (925) |
| Muslim | 56.1 (107) | 73.1 (402) | 45.7 (105) |
| Christian | n=38 | 96.5 (57) | n=11 |
| Other (Sikh, Buddhist, Jain, …) | 55.0 (100) | 79.1 (191) | 52.8 (74) |
| Caste | |||
| SC | — | 76.2 (697) | 58.4 (223) |
| ST | — | 76.8 (228) | 58.4 (59) |
| OBC | — | 69.7 (1452) | 52.3 (474) |
| General / other | — | 69.9 (1131) | 48.0 (361) |
| Cells: estimate (unweighted n). Cells with n < 50 are not estimated. | |||
plot_dumbbell("state_support", waves = c("W5", "W6", "W7"),
title = "India: state supporters by group", xlab = "% state supporters")Do the two kinds of question go together?
The first two questions ask what the government should do. The other three ask what democracy must guarantee. The table gives the correlation between the number of “yes” answers on each pair (0–2 and 0–3). Regions: mean of country correlations.
ss_cor <- wvs |>
filter(wave %in% c("W5", "W6", "W7")) |>
group_by(wave, country_code, macro_region) |>
filter(sum(!is.na(ss_policy_n) & !is.na(ss_essential_n)) > 100) |>
summarise(r = cor(ss_policy_n, ss_essential_n, use = "complete.obs"), .groups = "drop")
ss_cor |>
group_by(wave, macro_region) |>
summarise(r = mean(r), .groups = "drop") |>
mutate(macro_region = factor(macro_region, focus_regions), col = wave_labels[wave]) |>
select(Region = macro_region, col, r) |>
pivot_wider(names_from = col, values_from = r) |>
arrange(Region) |>
kable(digits = 2, caption = "Correlation: 'government should' questions vs 'democracy must' questions") |>
styled() |>
footnote(general = "2006 'democracy must' count uses two questions (state makes incomes equal not asked).",
general_title = "")| Region | W5 (2004–09) | W6 (2010–16) | W7 (2017–23) |
|---|---|---|---|
| India | 0.03 | 0.24 | 0.09 |
| Western Europe & Offshoots | 0.14 | 0.27 | 0.28 |
| Middle East & North Africa | 0.11 | 0.18 | 0.06 |
| 2006 ‘democracy must’ count uses two questions (state makes incomes equal not asked). |
State support and the conception of democracy
cross_tab("state_support", "State supporters (%), by conception of democracy")| India 2006 | India 2012 | India 2023 | W. Europe 2023 | MENA 2023 | |
|---|---|---|---|---|---|
| Liberal (maximalist) | 43.7 (309) | 85.6 (1056) | 55.1 (293) | 48.5 (12333) | 65.4 (2746) |
| Minimalist (electoral) | 49.7 (461) | 77.6 (1255) | 60.6 (441) | 47.0 (2656) | 69.0 (3961) |
| Illiberal | 38.4 (463) | 53.9 (1252) | 39.2 (359) | 25.6 (2803) | 43.5 (4500) |
| Cells: % (unweighted n). Cells with n < 50 are not estimated. |
What this shows.
- India’s state support rose and then fell back. State supporters (yes to at least 3) were 46% in 2006, 72% in 2012 and 52% in 2023. India went from 4th of 22 countries in 2012 to 12th of 21 in 2023. Western Europe rose steadily (32% → 42% → 45%); MENA stayed near 54–58%.
- The 2023 fall comes from the “government should” questions. “Government should take more responsibility” fell from 67% to 40%, and “incomes should be made more equal” from 71% to 21% (19th of 21 countries). The “democracy must” questions held up: tax the rich 65% → 69%, unemployment aid 69% → 66%, equal incomes 64% → 55%. The 2023 swing on the two economic scales also appears in the WVS’s own published tables, but is large enough to need checking against other surveys.
- The two kinds of question are only loosely linked. In India the correlation was 0.03 in 2006, 0.24 in 2012 and 0.09 in 2023; Western Europe 0.14–0.28, MENA 0.06–0.18. Read the category alongside the separate questions.
- State support goes with a democratic conception, liberal or minimalist. In India in 2023, 55% of liberals and 61% of minimalists were state supporters, against 39% of illiberals (2012: 86%, 78% and 54%). The same holds in Western Europe (49%, 47%, 26%) and MENA (65%, 69%, 44%). Wanting democracy and wanting a supportive state go together; illiberals want less of both.
- Within India, 2023, support is highest among the marginalised. 58% of SC and ST respondents were state supporters, against 48% of general castes. INC voters’ support fell furthest (79% → 51%), and is now close to BJP voters (48%).
5. Does it differ across Indian states?
The question. Fukuyama stresses that governance varies within countries. Do the patterns above hold across Indian states?
What is calculated. The main measures from sections 1 to 3 by state, for 2001 (18 states), 2012 (17 states) and 2023 (8 states). States with fewer than 150 respondents are marked; treat them as indicative only.
tribble(
~Measure, ~`Built from`, ~Section,
"Trust in the state / in checking institutions, and the gap", "8 confidence questions", "1",
"Strong leader good; experts good", "Two ways-of-governing questions (1–2 = good)", "3",
"Want both (governance and accountability)", "Yes on at least half of the delivery and of the accountability questions", "2") |>
kable(caption = "Measures used in this section (questions as listed in each section)") |>
styled()| Measure | Built from | Section |
|---|---|---|
| Trust in the state / in checking institutions, and the gap | 8 confidence questions | 1 |
| Strong leader good; experts good | Two ways-of-governing questions (1–2 = good) | 3 |
| Want both (governance and accountability) | Yes on at least half of the delivery and of the accountability questions | 2 |
state_tab <- function(w) {
india |>
filter(wave == w, !is.na(in_state)) |>
group_by(State = in_state) |>
summarise(n = n(),
`Trust in state %` = 100 * weighted.mean(state_trust, weight, na.rm = TRUE),
`Trust in checking %` = 100 * weighted.mean(check_trust, weight, na.rm = TRUE),
`Gap (pts)` = `Trust in state %` - `Trust in checking %`,
`Strong leader good %`= 100 * weighted.mean(reg_strongleader_good, weight, na.rm = TRUE),
`Experts good %` = 100 * weighted.mean(reg_experts_good, weight, na.rm = TRUE),
`Want both %` = 100 * weighted.mean(g_both, weight, na.rm = TRUE),
.groups = "drop") |>
arrange(desc(`Gap (pts)`))
}
for (w in c("W4", "W6", "W7")) {
tb <- state_tab(w)
print(tb |>
kable(digits = 1, caption = paste0("India ", india_years[w], ": measures by state (sorted by trust gap)")) |>
styled() |>
row_spec(which(tb$n < 150), color = "grey"))
cat("
")
}| State | n | Trust in state % | Trust in checking % | Gap (pts) | Strong leader good % | Experts good % | Want both % |
|---|---|---|---|---|---|---|---|
| Chhattisgarh | 46 | 61.1 | 42.9 | 18.3 | 100.0 | 70.0 | 42.9 |
| Madhya Pradesh | 115 | 56.2 | 40.7 | 15.6 | 86.4 | 72.9 | 61.8 |
| Kerala | 71 | 63.0 | 47.5 | 15.5 | 16.9 | 21.8 | 65.0 |
| Karnataka | 112 | 57.3 | 41.8 | 15.5 | 100.0 | 98.4 | 24.4 |
| Tamil Nadu | 129 | 87.8 | 73.7 | 14.1 | 48.7 | 71.3 | 46.9 |
| Jharkhand | 65 | 56.9 | 43.8 | 13.2 | 95.2 | 95.0 | 37.0 |
| Delhi | 43 | 45.0 | 33.3 | 11.7 | 69.2 | 91.9 | 25.7 |
| Odisha | 68 | 63.7 | 52.6 | 11.1 | 6.2 | 72.4 | 51.3 |
| Assam | 61 | 29.5 | 21.8 | 7.7 | 50.0 | 47.7 | 27.3 |
| Rajasthan | 102 | 76.9 | 70.0 | 7.0 | 92.6 | 96.2 | 34.7 |
| Andhra Pradesh | 152 | 57.5 | 50.8 | 6.7 | 37.1 | 45.0 | 48.3 |
| Uttar Pradesh | 328 | 58.0 | 55.1 | 2.9 | 76.9 | 61.6 | 49.5 |
| Maharashtra | 194 | 60.6 | 57.8 | 2.8 | 62.7 | 70.5 | 41.5 |
| Haryana | 50 | 40.4 | 38.5 | 1.9 | 100.0 | 100.0 | 41.2 |
| West Bengal | 174 | 59.2 | 57.6 | 1.6 | 29.3 | 64.1 | 12.7 |
| Gujarat | 96 | 58.7 | 58.6 | 0.1 | 17.9 | 69.1 | 40.4 |
| Bihar | 147 | 42.5 | 48.7 | -6.1 | 64.9 | 75.6 | 62.6 |
| Punjab | 49 | 59.5 | 75.7 | -16.2 | 62.5 | 36.2 | 81.2 |
| State | n | Trust in state % | Trust in checking % | Gap (pts) | Strong leader good % | Experts good % | Want both % |
|---|---|---|---|---|---|---|---|
| Unlabelled (W6 code 356019) | 64 | 52.0 | 29.9 | 22.1 | 93.5 | 96.8 | 33.3 |
| Rajasthan | 278 | 63.5 | 50.6 | 12.9 | 92.9 | 86.7 | 33.5 |
| Punjab | 151 | 59.7 | 47.1 | 12.6 | 79.8 | 73.0 | 16.4 |
| Gujarat | 259 | 60.1 | 49.9 | 10.3 | 79.8 | 52.2 | 49.2 |
| Bihar | 332 | 58.0 | 48.2 | 9.8 | 76.0 | 71.0 | 14.5 |
| Madhya Pradesh | 256 | 55.9 | 48.0 | 7.9 | 88.5 | 81.6 | 19.2 |
| Chhattisgarh | 99 | 66.2 | 58.4 | 7.7 | 98.3 | 93.0 | 46.7 |
| Odisha | 355 | 73.5 | 66.1 | 7.3 | 24.0 | 97.0 | 75.5 |
| Karnataka | 239 | 60.6 | 55.2 | 5.4 | 42.9 | 48.3 | 53.2 |
| Uttar Pradesh | 596 | 59.8 | 54.5 | 5.3 | 93.6 | 90.3 | 27.8 |
| Maharashtra | 280 | 73.4 | 68.2 | 5.2 | 53.2 | 56.1 | 28.7 |
| Andhra Pradesh | 307 | 68.3 | 69.1 | -0.8 | 39.3 | 47.6 | 34.9 |
| Haryana | 122 | 71.4 | 72.5 | -1.2 | 91.1 | 69.9 | 15.8 |
| West Bengal | 294 | 54.5 | 58.8 | -4.3 | 25.8 | 62.0 | 5.0 |
| Kerala | 192 | 78.5 | 82.9 | -4.4 | 7.9 | 36.6 | 50.8 |
| Delhi | 94 | 41.8 | 49.1 | -7.3 | 75.0 | 64.2 | 28.3 |
| Jharkhand | 160 | 44.2 | 59.4 | -15.2 | 96.6 | 83.8 | 72.3 |
| State | n | Trust in state % | Trust in checking % | Gap (pts) | Strong leader good % | Experts good % | Want both % |
|---|---|---|---|---|---|---|---|
| Telangana | 170 | 71.4 | 52.2 | 19.1 | 64.8 | 75.5 | 28.0 |
| Delhi | 187 | 83.9 | 70.0 | 13.9 | 88.2 | 76.8 | 33.1 |
| Bihar | 198 | 77.4 | 64.7 | 12.7 | 74.2 | 58.4 | 39.5 |
| Haryana | 95 | 74.0 | 61.7 | 12.3 | 76.6 | 67.6 | 31.9 |
| Punjab | 100 | 44.5 | 35.0 | 9.5 | 31.8 | 81.8 | 46.3 |
| Uttar Pradesh | 292 | 78.1 | 70.6 | 7.4 | 94.3 | 80.8 | 37.0 |
| West Bengal | 169 | 74.7 | 67.7 | 7.0 | 27.2 | 60.2 | 20.8 |
| Maharashtra | 160 | 76.8 | 70.9 | 5.8 | 71.0 | 82.0 | 28.2 |
p <- india |>
filter(wave %in% c("W4", "W6", "W7"), !is.na(in_state), !grepl("Unlabelled", in_state)) |>
group_by(wave, in_state) |>
summarise(n = n(), state_trust = 100 * weighted.mean(state_trust, weight, na.rm = TRUE),
strongleader = 100 * weighted.mean(reg_strongleader_good, weight, na.rm = TRUE), .groups = "drop") |>
filter(n >= 80) |>
mutate(year = factor(india_years[wave], india_years)) |>
ggplot(aes(state_trust, strongleader)) +
geom_point_interactive(aes(data_id = paste(in_state, year),
tooltip = sprintf("%s, %s<br>Trust in state: %.0f%%<br>Strong leader good: %.0f%%<br>n = %d",
in_state, year, state_trust, strongleader, n)),
shape = 21, fill = region_cols["India"], colour = "white", size = 3.4, stroke = 0.8) +
geom_text_repel(aes(label = in_state), size = 2.7, colour = ink2, max.overlaps = 30, seed = 3) +
facet_wrap(~ year, nrow = 1) +
scale_x_continuous(breaks = c(40, 60, 80)) +
labs(title = "Indian states: trust in the state vs support for a strong leader",
x = "Trust in the state (%)", y = "Strong leader is good (%)",
caption = "States with at least 80 respondents.") +
theme_wvs(10) + theme(panel.spacing.x = unit(1.5, "lines"))
show_plot(p, 11, 5)What this shows.
- States differ enormously, more than India differs from the two regions. In 2012 support for a strong leader ranges from 8% in Kerala and 24–26% in Odisha and West Bengal to over 90% in Uttar Pradesh, Rajasthan, Jharkhand, Chhattisgarh and Haryana.
- National India figures therefore depend heavily on which states are sampled. This is a further reason to read the 8-state 2023 sample with care.
- The state labelled “Unlabelled (W6 code 356019)” has no name in the codebook and only 64 respondents; ignore it.
Robustness: the seven states surveyed in every wave
head_items <- c(state_trust = "Trust in the state", check_trust = "Trust in checking institutions",
trust_gap = "Trust gap (pts)", g_both = "Want both", g_govfirst = "Governance-first")
bind_rows(
india_trend(head_items) |> mutate(sample = "All states"),
india_trend(head_items, data = filter(india, in_state %in% common_states)) |> mutate(sample = "Seven common states")
) |>
mutate(cell = ifelse(is.na(est), "—", sprintf("%.1f", est * 100)), year = india_years[wave]) |>
select(Item = item, Sample = sample, year, cell) |>
pivot_wider(names_from = year, values_from = cell) |>
arrange(match(Item, head_items)) |>
kable(caption = "India: key measures, all states vs Bihar, Delhi, Haryana, Maharashtra, Punjab, UP, West Bengal",
align = c("l", "l", rep("c", 5))) |>
styled() |>
collapse_rows(columns = 1, valign = "top")| Item | Sample | 2001 | 2006 | 2012 | 2023 |
|---|---|---|---|---|---|
| Trust in the state | All states | 59.1 | 64.7 | 62.5 | 74.7 |
| Seven common states | 55.4 | 65.0 | 60.6 | 75.2 | |
| Trust in checking institutions | All states | 53.8 | 63.8 | 57.9 | 63.8 |
| Seven common states | 54.8 | 61.6 | 56.3 | 65.5 | |
| Trust gap (pts) | All states | 4.7 | 0.7 | 4.2 | 10.7 |
| Seven common states | 0.7 | 3.4 | 3.8 | 9.6 | |
| Want both | All states | 43.0 | 43.7 | 35.0 | 32.7 |
| Seven common states | 42.5 | 39.3 | 19.1 | 33.4 | |
| Governance-first | All states | 47.3 | 44.7 | 53.4 | 44.0 |
| Seven common states | 51.5 | 52.0 | 69.3 | 44.8 |