Follow the Money: Revenue Sources and Financing Gaps
source(here::here("code", "00_setup.R"))
# Main analysis data (candidate-level)
df <- readRDS(paths$analysis_full_rds)
# Raw TSE revenue source categories (saved by 02_load_finance_raw.R)
raw_cats <- read_csv(
paste0(paths$tables, "finance_revenue_source_categories.csv"),
show_col_types = FALSE
) %>%
filter(!is.na(DS_ORIGEM_RECEITA))
# Transaction-level finance data (for in-kind analysis)
receitas <- readRDS(paths$finance_trans_rds)# --- Pre-compute POOLED values for inline narrative text in Summary ---
# Overall stats
n_total <- nrow(df)
n_zero_rev <- sum(df$total_revenue == 0)
pct_zero_rev <- mean(df$total_revenue == 0)
median_rev_all <- median(df$total_revenue)
mean_rev_all <- mean(df$total_revenue)
# LGBTQ+ vs Non-LGBTQ+ comparison (pooled, for Summary)
lgbtq_stats <- df %>%
group_by(lgbtq_candidate) %>%
summarise(
n = n(),
mean_rev = mean(total_revenue),
median_rev = median(total_revenue),
pct_zero = mean(total_revenue == 0),
mean_pct_self = mean(pct_self, na.rm = TRUE),
mean_pct_party = mean(pct_party, na.rm = TRUE),
mean_pct_individual = mean(pct_individual, na.rm = TRUE),
mean_pct_crowdfunding = mean(pct_crowdfunding, na.rm = TRUE),
mean_pct_financial = mean(pct_financial, na.rm = TRUE),
mean_pct_inkind = mean(pct_inkind, na.rm = TRUE),
.groups = "drop"
)
lgbtq_median <- lgbtq_stats$median_rev[lgbtq_stats$lgbtq_candidate == TRUE]
nonlgbtq_median <- lgbtq_stats$median_rev[lgbtq_stats$lgbtq_candidate == FALSE]
lgbtq_mean <- lgbtq_stats$mean_rev[lgbtq_stats$lgbtq_candidate == TRUE]
nonlgbtq_mean <- lgbtq_stats$mean_rev[lgbtq_stats$lgbtq_candidate == FALSE]
median_ratio_overall <- lgbtq_median / nonlgbtq_median
mean_ratio_overall <- lgbtq_mean / nonlgbtq_mean
# Direction labels for inline use
median_direction <- if (lgbtq_median > nonlgbtq_median) "higher" else if (lgbtq_median < nonlgbtq_median) "lower" else "equal"
mean_direction <- if (lgbtq_mean > nonlgbtq_mean) "higher" else if (lgbtq_mean < nonlgbtq_mean) "lower" else "equal"
# LGBTQ+ funding composition
lgbtq_pct_party <- lgbtq_stats$mean_pct_party[lgbtq_stats$lgbtq_candidate == TRUE]
nonlgbtq_pct_party <- lgbtq_stats$mean_pct_party[lgbtq_stats$lgbtq_candidate == FALSE]
lgbtq_pct_individual <- lgbtq_stats$mean_pct_individual[lgbtq_stats$lgbtq_candidate == TRUE]
nonlgbtq_pct_individual <- lgbtq_stats$mean_pct_individual[lgbtq_stats$lgbtq_candidate == FALSE]
lgbtq_pct_self <- lgbtq_stats$mean_pct_self[lgbtq_stats$lgbtq_candidate == TRUE]
nonlgbtq_pct_self <- lgbtq_stats$mean_pct_self[lgbtq_stats$lgbtq_candidate == FALSE]
# Identity category stats (pooled, for Summary)
identity_stats <- df %>%
filter(lgbtq_candidate, lgbt_category != "Other LGBTQ+") %>%
group_by(lgbt_category) %>%
summarise(
mean_pct_party = mean(pct_party, na.rm = TRUE),
mean_pct_individual = mean(pct_individual, na.rm = TRUE),
median_rev = median(total_revenue),
.groups = "drop"
)
# Find which category has the highest party share
top_party_cat <- identity_stats %>% slice_max(mean_pct_party, n = 1) %>% pull(lgbt_category)
top_party_val <- identity_stats %>% slice_max(mean_pct_party, n = 1) %>% pull(mean_pct_party)
top_indiv_cat <- identity_stats %>% slice_max(mean_pct_individual, n = 1) %>% pull(lgbt_category)
top_indiv_val <- identity_stats %>% slice_max(mean_pct_individual, n = 1) %>% pull(mean_pct_individual)Campaign finance is the lifeblood of electoral competition. In Brazil’s municipal elections, candidates must register all revenue with the Tribunal Superior Eleitoral (TSE), creating a comprehensive public record of who funds whom. This chapter provides a transparent, ground-up analysis of campaign revenue — from raw source categories through aggregate patterns — with special attention to how LGBTQ+ candidates’ financial profiles differ from the broader candidate population.
We proceed in four stages: (1) documenting the raw data categories for full transparency, (2) describing general revenue patterns across all candidates, (3) comparing LGBTQ+ and non-LGBTQ+ financing, and (4) disaggregating by identity category and party.
Campaign revenue differs dramatically by position: city councilor campaigns are typically an order of magnitude smaller than mayoral campaigns. To avoid conflating these fundamentally different scales, the main comparison sections are presented separately for city councilors and mayors/vice-mayors using tabbed panels.
The TSE raw finance file classifies each transaction by DS_ORIGEM_RECEITA (revenue source). Before any analysis, we display the complete frequency table of these raw categories, along with our classification scheme. This is the foundation of all subsequent finance analysis.
# Add the classification mapping used in 02_load_finance_raw.R
raw_cats <- raw_cats %>%
mutate(
classification = case_when(
str_detect(DS_ORIGEM_RECEITA, "(?i)recursos pr") ~ "Self-funding",
str_detect(DS_ORIGEM_RECEITA, "(?i)partido pol") ~ "Party funding",
str_detect(DS_ORIGEM_RECEITA, "(?i)pessoas f") ~ "Individual donation",
str_detect(DS_ORIGEM_RECEITA, "(?i)financiamento coletivo") ~ "Crowdfunding",
str_detect(DS_ORIGEM_RECEITA, "(?i)outros candidatos") ~ "Other candidates",
str_detect(DS_ORIGEM_RECEITA, "(?i)internet") ~ "Online donations",
str_detect(DS_ORIGEM_RECEITA, "#NULO") ~ "Null/Unclassified",
TRUE ~ "Other"
),
total_brl = replace_na(total_brl, 0),
total_brl_fmt = format_brl(total_brl),
pct_fmt = replace_na(pct, "0.0%"),
n = replace_na(n, 0L)
)
raw_cats %>%
select(
`TSE Category` = DS_ORIGEM_RECEITA,
`N Transactions` = n,
`% of Trans.` = pct_fmt,
`Total (R$)` = total_brl_fmt,
`Our Label` = classification
) %>%
kable(align = c("l", "r", "r", "r", "l"))| TSE Category | N Transactions | % of Trans. | Total (R$) | Our Label |
|---|---|---|---|---|
| Recursos de outros candidatos | 669606 | 32.9% | R$ 260,338,476 | Other candidates |
| Recursos de pessoas físicas | 561708 | 27.6% | R$1,221,519,516 | Individual donation |
| Recursos de partido político | 377468 | 18.5% | R$5,104,749,805 | Party funding |
| Recursos próprios | 263827 | 12.9% | R$ 426,344,883 | Self-funding |
| Doações pela Internet | 90850 | 4.5% | R$ 6,457,032 | Online donations |
| #NULO | 66718 | 3.3% | R$ 0 | Null/Unclassified |
| Recursos de Financiamento Coletivo | 4324 | 0.2% | R$ 8,092,035 | Crowdfunding |
| Recursos de origens não identificadas | 2687 | 0.1% | R$ 1,071,403 | Other |
| Rendimentos de aplicações financeiras | 263 | 0.0% | R$ 115,533 | Other |
| Comercialização de Bens com OR | 109 | 0.0% | R$ 37,252 | Other |
| Comercialização de Bens com FEFC | 17 | 0.0% | R$ 18,289 | Other |
Every transaction in the TSE receitas file is classified using the mapping above. The raw Portuguese-language category names are preserved alongside our English labels so that any researcher can verify the mapping. The classification code is in code/02_load_finance_raw.R.
The measure total_revenue captures the sum of all campaign receipts registered with the TSE for each candidate, denominated in Brazilian reais (R$). This includes financial contributions (cash, PIX, bank transfers) and in-kind/estimated contributions (goods and services assigned a monetary value). A candidate with zero total revenue either received no contributions or did not report any.
render_general_revenue <- function(data, tab_name) {
# --- Revenue summary statistics ---
cat("### Revenue Summary Statistics\n\n")
revenue_stats <- data %>%
summarise(
N = format_n(n()),
`Zero Revenue` = format_n(sum(total_revenue == 0)),
`% Zero` = format_pct(mean(total_revenue == 0)),
Mean = format_brl(mean(total_revenue)),
SD = format_brl(sd(total_revenue)),
Median = format_brl(median(total_revenue)),
P25 = format_brl(quantile(total_revenue, 0.25)),
P75 = format_brl(quantile(total_revenue, 0.75)),
P99 = format_brl(quantile(total_revenue, 0.99)),
Max = format_brl(max(total_revenue))
)
revenue_stats %>%
pivot_longer(everything(), names_to = "Statistic", values_to = "Value") %>%
cat_kable(align = c("l", "r"))
# --- Revenue distribution histogram ---
cat("### Revenue Distribution (Log Scale)\n\n")
pos_rev <- data %>% filter(total_revenue > 0)
if (nrow(pos_rev) > 0) {
med_val <- median(pos_rev$total_revenue)
p_dist <- pos_rev %>%
ggplot(aes(x = log10(total_revenue))) +
geom_histogram(binwidth = 0.2, fill = "#3498DB", alpha = 0.7, color = "white") +
geom_vline(xintercept = log10(med_val),
linetype = "dashed", color = "#E74C3C", linewidth = 0.8) +
annotate("text",
x = log10(med_val) + 0.3, y = Inf, vjust = 2, hjust = 0,
label = paste0("Median = ", format_brl(med_val)),
color = "#E74C3C", fontface = "bold", size = 4) +
scale_x_continuous(
breaks = 0:7,
labels = c("R$1", "R$10", "R$100", "R$1K", "R$10K", "R$100K", "R$1M", "R$10M")
) +
labs(
x = "Total Revenue (log scale)",
y = "Number of Candidates",
title = paste0("Campaign Revenue Distribution (", tab_name, ")"),
subtitle = "Among candidates with positive revenue; dashed line = median",
caption = "Excludes candidates with zero reported revenue."
)
cat_plot(p_dist, paste0("05-revenue-dist-", pos_suffix(tab_name)))
}
# --- Revenue composition ---
cat("### Revenue Composition by Source\n\n")
comp <- data %>%
summarise(
`Self-funding` = sum(self_funding_amt, na.rm = TRUE),
`Party funding` = sum(party_funding_amt, na.rm = TRUE),
`Individual donors` = sum(individual_funding_amt, na.rm = TRUE),
`Crowdfunding` = sum(crowdfunding_amt, na.rm = TRUE),
`Other` = sum(total_revenue, na.rm = TRUE) -
sum(self_funding_amt, na.rm = TRUE) -
sum(party_funding_amt, na.rm = TRUE) -
sum(individual_funding_amt, na.rm = TRUE) -
sum(crowdfunding_amt, na.rm = TRUE)
) %>%
pivot_longer(everything(), names_to = "source", values_to = "total") %>%
mutate(
pct = total / sum(total),
source = factor(source,
levels = c("Party funding", "Self-funding", "Individual donors",
"Crowdfunding", "Other"))
)
p_comp <- ggplot(comp, aes(x = "", y = pct, fill = source)) +
geom_col(width = 0.6, alpha = 0.9) +
geom_text(aes(label = paste0(format_pct(pct), "\n", format_brl(total))),
position = position_stack(vjust = 0.5), size = 3.5, color = "white",
fontface = "bold") +
coord_flip() +
scale_fill_brewer(palette = "Set2", name = "Funding Source") +
scale_y_continuous(labels = percent) +
labs(
x = NULL,
y = "Share of Total Revenue",
title = paste0("Where the Money Comes From (", tab_name, ")"),
subtitle = "Aggregate revenue composition across all candidates"
)
cat_plot(p_comp, paste0("05-composition-overall-", pos_suffix(tab_name)))
}
render_position_tabset(render_general_revenue, df)| Statistic | Value |
|---|---|
| N | 432,005 |
| Zero Revenue | 69,333 |
| % Zero | 16.0% |
| Mean | R$7,348 |
| SD | R$34,882 |
| Median | R$1,769 |
| P25 | R$355 |
| P75 | R$5,337 |
| P99 | R$94,264 |
| Max | R$4,288,678 |
| Statistic | Value |
|---|---|
| N | 31,596 |
| Zero Revenue | 16,320 |
| % Zero | 51.7% |
| Mean | R$142,697 |
| SD | R$1,051,994 |
| Median | R$0 |
| P25 | R$0 |
| P75 | R$112,528 |
| P99 | R$2,012,627 |
| Max | R$81,590,273 |
This tab pools city councilors (proportional representation) and mayors/vice-mayors (plurality). Position-specific results in the other tabs may be more informative.
| Statistic | Value |
|---|---|
| N | 463,601 |
| Zero Revenue | 85,653 |
| % Zero | 18.5% |
| Mean | R$16,573 |
| SD | R$278,783 |
| Median | R$1,752 |
| P25 | R$299 |
| P75 | R$5,792 |
| P99 | R$204,681 |
| Max | R$81,590,273 |
render_lgbtq_revenue <- function(data, tab_name) {
# --- Side-by-side comparison table ---
cat("### Revenue Comparison\n\n")
revenue_comparison <- data %>%
mutate(group = if_else(lgbtq_candidate, "LGBTQ+", "Non-LGBTQ+")) %>%
group_by(group) %>%
summarise(
N = format_n(n()),
`% Zero Rev.` = format_pct(mean(total_revenue == 0)),
`Mean Rev.` = format_brl(mean(total_revenue)),
`Median Rev.` = format_brl(median(total_revenue)),
`SD Rev.` = format_brl(sd(total_revenue)),
`Mean Trans.` = as.character(round(mean(n_transactions), 1)),
`Mean Donors` = as.character(round(mean(n_unique_donors), 1)),
`% Self-fund` = format_pct100(mean(pct_self)),
`% Party` = format_pct100(mean(pct_party)),
`% Individual` = format_pct100(mean(pct_individual)),
`% Crowdfund` = format_pct100(mean(pct_crowdfunding)),
.groups = "drop"
)
revenue_comparison %>%
pivot_longer(-group, names_to = "Metric", values_to = "value") %>%
pivot_wider(names_from = group, values_from = value) %>%
cat_kable(align = c("l", "r", "r"))
# Compute position-specific stats for callout
pos_stats <- data %>%
group_by(lgbtq_candidate) %>%
summarise(
median_rev = median(total_revenue),
mean_rev = mean(total_revenue),
.groups = "drop"
)
lgbtq_med <- pos_stats$median_rev[pos_stats$lgbtq_candidate == TRUE]
non_med <- pos_stats$median_rev[pos_stats$lgbtq_candidate == FALSE]
lgbtq_mn <- pos_stats$mean_rev[pos_stats$lgbtq_candidate == TRUE]
non_mn <- pos_stats$mean_rev[pos_stats$lgbtq_candidate == FALSE]
med_ratio <- if (non_med > 0) lgbtq_med / non_med else NA_real_
mn_ratio <- if (non_mn > 0) lgbtq_mn / non_mn else NA_real_
cat("::: {.callout-note}\n")
cat("## Revenue Comparison\n")
cat(sprintf("In this tab (%s), the median revenue for LGBTQ+ candidates is %s vs. %s for non-LGBTQ+ (ratio: %.2f). ",
tab_name, format_brl(lgbtq_med), format_brl(non_med),
if (!is.na(med_ratio)) med_ratio else 0))
cat(sprintf("At the mean: %s vs. %s (ratio: %.2f).\n",
format_brl(lgbtq_mn), format_brl(non_mn),
if (!is.na(mn_ratio)) mn_ratio else 0))
cat(":::\n\n")
# --- Revenue distribution comparison ---
cat("### Revenue Distribution\n\n")
pos_rev <- data %>% filter(total_revenue > 0)
if (nrow(pos_rev) > 0) {
p_density <- pos_rev %>%
mutate(group = if_else(lgbtq_candidate, "LGBTQ+", "Non-LGBTQ+")) %>%
ggplot(aes(x = log10(total_revenue), fill = group, color = group)) +
geom_density(alpha = 0.3, linewidth = 0.8) +
scale_fill_manual(values = pal_lgbtq, name = NULL) +
scale_color_manual(values = pal_lgbtq, name = NULL) +
scale_x_continuous(
breaks = 0:7,
labels = c("R$1", "R$10", "R$100", "R$1K", "R$10K", "R$100K", "R$1M", "R$10M")
) +
labs(
x = "Total Revenue (log scale)",
y = "Density",
title = paste0("Revenue Distribution (", tab_name, ")"),
subtitle = "Density comparison among candidates with positive revenue"
)
cat_plot(p_density, paste0("05-density-revenue-", pos_suffix(tab_name)))
# Box plot
p_box <- pos_rev %>%
mutate(group = if_else(lgbtq_candidate, "LGBTQ+", "Non-LGBTQ+")) %>%
ggplot(aes(x = group, y = total_revenue, fill = group)) +
geom_boxplot(alpha = 0.7, outlier.alpha = 0.1, outlier.size = 0.5) +
scale_y_log10(labels = label_dollar(prefix = "R$", big.mark = ",")) +
scale_fill_manual(values = pal_lgbtq, guide = "none") +
labs(
x = NULL,
y = "Total Revenue (log scale)",
title = paste0("Campaign Revenue by LGBTQ+ Status (", tab_name, ")"),
subtitle = "Among candidates with positive revenue"
)
cat_plot(p_box, paste0("05-box-revenue-", pos_suffix(tab_name)))
}
# --- Funding composition ---
cat("### Funding Source Composition\n\n")
p_comp <- data %>%
mutate(group = if_else(lgbtq_candidate, "LGBTQ+", "Non-LGBTQ+")) %>%
group_by(group) %>%
summarise(
`Self-funding` = sum(self_funding_amt, na.rm = TRUE),
`Party funding` = sum(party_funding_amt, na.rm = TRUE),
`Individual donors` = sum(individual_funding_amt, na.rm = TRUE),
`Crowdfunding` = sum(crowdfunding_amt, na.rm = TRUE),
.groups = "drop"
) %>%
pivot_longer(-group, names_to = "source", values_to = "total") %>%
group_by(group) %>%
mutate(pct = total / sum(total)) %>%
ungroup() %>%
mutate(source = factor(source,
levels = c("Party funding", "Self-funding",
"Individual donors", "Crowdfunding"))) %>%
ggplot(aes(x = group, y = pct, fill = source)) +
geom_col(alpha = 0.9, width = 0.6) +
geom_text(aes(label = format_pct(pct)),
position = position_stack(vjust = 0.5), size = 3.5, color = "white",
fontface = "bold") +
scale_fill_brewer(palette = "Set2", name = "Funding Source") +
scale_y_continuous(labels = percent) +
labs(
x = NULL,
y = "Share of Total Revenue",
title = paste0("Funding Portfolios (", tab_name, ")"),
subtitle = "Revenue composition by source for LGBTQ+ vs Non-LGBTQ+ candidates"
)
cat_plot(p_comp, paste0("05-composition-lgbtq-", pos_suffix(tab_name)))
}
render_position_tabset(render_lgbtq_revenue, df)| Metric | LGBTQ+ | Non-LGBTQ+ |
|---|---|---|
| N | 3,043 | 428,962 |
| % Zero Rev. | 12.2% | 16.1% |
| Mean Rev. | R$18,672 | R$7,268 |
| Median Rev. | R$3,785 | R$1,759 |
| SD Rev. | R$56,363 | R$34,669 |
| Mean Trans. | 5.3 | 4 |
| Mean Donors | 3 | 2.2 |
| % Self-fund | 10.5% | 15.3% |
| % Party | 48.0% | 26.6% |
| % Individual | 12.8% | 17.9% |
| % Crowdfund | 0.4% | 0.1% |
In this tab (City Councilors), the median revenue for LGBTQ+ candidates is R$3,785 vs. R$1,759 for non-LGBTQ+ (ratio: 2.15). At the mean: R$18,672 vs. R$7,268 (ratio: 2.57).
| Metric | LGBTQ+ | Non-LGBTQ+ |
|---|---|---|
| N | 91 | 31,505 |
| % Zero Rev. | 58.2% | 51.6% |
| Mean Rev. | R$202,403 | R$142,524 |
| Median Rev. | R$0 | R$0 |
| SD Rev. | R$1,013,122 | R$1,052,115 |
| Mean Trans. | 5.8 | 9.9 |
| Mean Donors | 3.8 | 7.6 |
| % Self-fund | 1.1% | 4.3% |
| % Party | 35.5% | 35.0% |
| % Individual | 4.8% | 8.9% |
| % Crowdfund | 0.3% | 0.0% |
In this tab (Mayors & Vice-Mayors), the median revenue for LGBTQ+ candidates is R$0 vs. R$0 for non-LGBTQ+ (ratio: 0.00). At the mean: R$202,403 vs. R$142,524 (ratio: 1.42).
This tab pools city councilors (proportional representation) and mayors/vice-mayors (plurality). Position-specific results in the other tabs may be more informative.
| Metric | LGBTQ+ | Non-LGBTQ+ |
|---|---|---|
| N | 3,134 | 460,467 |
| % Zero Rev. | 13.6% | 18.5% |
| Mean Rev. | R$24,007 | R$16,522 |
| Median Rev. | R$3,705 | R$1,746 |
| SD Rev. | R$183,090 | R$279,321 |
| Mean Trans. | 5.3 | 4.4 |
| Mean Donors | 3 | 2.5 |
| % Self-fund | 10.2% | 14.6% |
| % Party | 47.6% | 27.2% |
| % Individual | 12.6% | 17.3% |
| % Crowdfund | 0.4% | 0.1% |
In this tab (All Candidates), the median revenue for LGBTQ+ candidates is R$3,705 vs. R$1,746 for non-LGBTQ+ (ratio: 2.12). At the mean: R$24,007 vs. R$16,522 (ratio: 1.45).
render_identity_revenue <- function(data, tab_name) {
simplified <- use_simplified(data)
# --- Revenue summary by category ---
cat("### Revenue Summary by Category\n\n")
identity_revenue <- data %>%
filter(lgbtq_candidate, lgbt_category != "Other LGBTQ+") %>%
group_by(lgbt_category) %>%
summarise(
N = format_n(n()),
`% Zero` = format_pct(mean(total_revenue == 0)),
`Mean Rev.` = format_brl(mean(total_revenue)),
`Median Rev.` = format_brl(median(total_revenue)),
`Mean Trans.` = as.character(round(mean(n_transactions), 1)),
`Mean Donors` = as.character(round(mean(n_unique_donors), 1)),
.groups = "drop"
) %>%
rename(Category = lgbt_category)
# Non-LGBTQ+ reference row
non_lgbtq_row <- data %>%
filter(!lgbtq_candidate) %>%
summarise(
Category = "Non-LGBTQ+ (ref.)",
N = format_n(n()),
`% Zero` = format_pct(mean(total_revenue == 0)),
`Mean Rev.` = format_brl(mean(total_revenue)),
`Median Rev.` = format_brl(median(total_revenue)),
`Mean Trans.` = as.character(round(mean(n_transactions), 1)),
`Mean Donors` = as.character(round(mean(n_unique_donors), 1))
)
bind_rows(identity_revenue, non_lgbtq_row) %>%
cat_kable(align = c("l", "r", "r", "r", "r", "r", "r"))
# --- Violin plot ---
cat("### Revenue Distribution by Category\n\n")
pos_rev <- data %>% filter(total_revenue > 0)
if (nrow(pos_rev) > 0) {
plot_data <- pos_rev %>%
mutate(
plot_category = if_else(lgbtq_candidate, as.character(lgbt_category), "Non-LGBTQ+"),
plot_category = factor(plot_category,
levels = c("Non-LGBTQ+", "Gay", "Lesbian", "Bisexual+",
"Trans", "Asexual"))
) %>%
filter(!is.na(plot_category))
if (simplified) {
# Small N: boxplot + jitter for LGBTQ+ categories
p_violin <- plot_data %>%
ggplot(aes(x = plot_category, y = total_revenue, fill = plot_category)) +
geom_boxplot(alpha = 0.7, outlier.shape = NA, width = 0.5) +
geom_jitter(data = . %>% filter(plot_category != "Non-LGBTQ+"),
alpha = 0.4, width = 0.15, size = 1) +
scale_y_log10(labels = label_dollar(prefix = "R$", big.mark = ",")) +
scale_fill_manual(values = pal_identity, guide = "none") +
labs(
x = NULL, y = "Total Revenue (log scale)",
title = paste0("Revenue by Identity (", tab_name, ")"),
subtitle = "Boxplot with individual LGBTQ+ observations"
) +
theme(axis.text.x = element_text(angle = 30, hjust = 1))
} else {
p_violin <- plot_data %>%
ggplot(aes(x = plot_category, y = total_revenue, fill = plot_category)) +
geom_violin(alpha = 0.7, draw_quantiles = c(0.25, 0.5, 0.75)) +
scale_y_log10(labels = label_dollar(prefix = "R$", big.mark = ",")) +
scale_fill_manual(values = pal_identity, guide = "none") +
labs(
x = NULL, y = "Total Revenue (log scale)",
title = paste0("Revenue by Identity (", tab_name, ")"),
subtitle = "Violin plots with quartile lines; among candidates with positive revenue"
) +
theme(axis.text.x = element_text(angle = 30, hjust = 1))
}
cat_plot(p_violin, paste0("05-violin-identity-", pos_suffix(tab_name)))
}
# --- Funding composition by identity ---
cat("### Funding Composition by Category\n\n")
comp_data <- data %>%
mutate(
plot_category = if_else(lgbtq_candidate, as.character(lgbt_category), "Non-LGBTQ+"),
plot_category = factor(plot_category,
levels = c("Non-LGBTQ+", "Gay", "Lesbian", "Bisexual+",
"Trans", "Asexual"))
) %>%
filter(!is.na(plot_category)) %>%
group_by(plot_category) %>%
summarise(
`Self-funding` = sum(self_funding_amt, na.rm = TRUE),
`Party funding` = sum(party_funding_amt, na.rm = TRUE),
`Individual donors` = sum(individual_funding_amt, na.rm = TRUE),
`Crowdfunding` = sum(crowdfunding_amt, na.rm = TRUE),
.groups = "drop"
) %>%
pivot_longer(-plot_category, names_to = "source", values_to = "total") %>%
group_by(plot_category) %>%
mutate(pct = total / sum(total)) %>%
ungroup() %>%
mutate(source = factor(source,
levels = c("Party funding", "Self-funding",
"Individual donors", "Crowdfunding")))
p_comp_id <- comp_data %>%
ggplot(aes(x = plot_category, y = pct, fill = source)) +
geom_col(alpha = 0.9) +
geom_text(aes(label = if_else(pct > 0.05, format_pct(pct), "")),
position = position_stack(vjust = 0.5), size = 3, color = "white",
fontface = "bold") +
scale_fill_brewer(palette = "Set2", name = "Funding Source") +
scale_y_continuous(labels = percent) +
labs(
x = NULL,
y = "Share of Total Revenue",
title = paste0("Funding Portfolio by Identity (", tab_name, ")"),
subtitle = "Revenue composition by funding source across identity categories"
) +
theme(axis.text.x = element_text(angle = 30, hjust = 1))
cat_plot(p_comp_id, paste0("05-composition-identity-", pos_suffix(tab_name)), height = 8)
# Position-specific within-group callout
id_stats <- data %>%
filter(lgbtq_candidate, lgbt_category != "Other LGBTQ+") %>%
group_by(lgbt_category) %>%
summarise(
mean_pct_party = mean(pct_party, na.rm = TRUE),
mean_pct_individual = mean(pct_individual, na.rm = TRUE),
.groups = "drop"
)
if (nrow(id_stats) > 0) {
top_p <- id_stats %>% slice_max(mean_pct_party, n = 1)
top_i <- id_stats %>% slice_max(mean_pct_individual, n = 1)
cat("::: {.callout-note}\n")
cat("## Within-Group Differences\n")
cat(sprintf("In this tab (%s), %s candidates have the highest average party funding share (%s), ",
tab_name, top_p$lgbt_category, format_pct100(top_p$mean_pct_party)))
cat(sprintf("while %s candidates have the highest average individual donor share (%s).\n",
top_i$lgbt_category, format_pct100(top_i$mean_pct_individual)))
cat(":::\n\n")
}
}
render_position_tabset(render_identity_revenue, df)| Category | N | % Zero | Mean Rev. | Median Rev. | Mean Trans. | Mean Donors |
|---|---|---|---|---|---|---|
| Gay | 1,034 | 11.2% | R$13,646 | R$3,496 | 5.6 | 3.3 |
| Lesbian | 670 | 11.5% | R$19,009 | R$4,450 | 4.9 | 2.7 |
| Bisexual+ | 538 | 9.1% | R$34,739 | R$6,627 | 6.7 | 3.7 |
| Trans | 610 | 15.6% | R$16,417 | R$3,000 | 4.5 | 2.4 |
| Asexual | 188 | 18.6% | R$6,281 | R$1,423 | 3.1 | 1.8 |
| Non-LGBTQ+ (ref.) | 428,962 | 16.1% | R$7,268 | R$1,759 | 4 | 2.2 |
In this tab (City Councilors), Bisexual+ candidates have the highest average party funding share (59.8%), while Asexual candidates have the highest average individual donor share (16.1%).
| Category | N | % Zero | Mean Rev. | Median Rev. | Mean Trans. | Mean Donors |
|---|---|---|---|---|---|---|
| Gay | 43 | 51.2% | R$90,946 | R$0 | 6.2 | 3.7 |
| Lesbian | 11 | 90.9% | R$1,693 | R$0 | 0.2 | 0.1 |
| Bisexual+ | 26 | 57.7% | R$218,384 | R$0 | 5.5 | 3.8 |
| Trans | 4 | 25.0% | R$2,146,515 | R$40,546 | 26.2 | 20 |
| Asexual | 7 | 71.4% | R$32,190 | R$0 | 1.4 | 1.1 |
| Non-LGBTQ+ (ref.) | 31,505 | 51.6% | R$142,524 | R$0 | 9.9 | 7.6 |
In this tab (Mayors & Vice-Mayors), Trans candidates have the highest average party funding share (72.6%), while Gay candidates have the highest average individual donor share (9.3%).
This tab pools city councilors (proportional representation) and mayors/vice-mayors (plurality). Position-specific results in the other tabs may be more informative.
| Category | N | % Zero | Mean Rev. | Median Rev. | Mean Trans. | Mean Donors |
|---|---|---|---|---|---|---|
| Gay | 1,077 | 12.8% | R$16,732 | R$3,494 | 5.7 | 3.3 |
| Lesbian | 681 | 12.8% | R$18,729 | R$4,104 | 4.9 | 2.6 |
| Bisexual+ | 564 | 11.3% | R$43,205 | R$6,460 | 6.6 | 3.7 |
| Trans | 614 | 15.6% | R$30,294 | R$3,018 | 4.7 | 2.5 |
| Asexual | 195 | 20.5% | R$7,211 | R$1,330 | 3 | 1.8 |
| Non-LGBTQ+ (ref.) | 460,467 | 18.5% | R$16,522 | R$1,746 | 4.4 | 2.5 |
In this tab (All Candidates), Bisexual+ candidates have the highest average party funding share (58.9%), while Asexual candidates have the highest average individual donor share (15.6%).
LGBTQ+ candidates tend to run in larger, more urbanized municipalities (see Chapter 4). Since larger cities have more expensive campaigns, a naive revenue comparison may overstate LGBTQ+ fundraising capacity. This section compares revenue within population brackets to reveal whether the aggregate pattern holds when municipality size is accounted for.
rev_by_pop <- df %>%
filter(!is.na(pop_bracket), total_revenue > 0) %>%
group_by(pop_bracket, lgbtq_label) %>%
summarise(
N = n(),
median_rev = median(total_revenue, na.rm = TRUE),
mean_rev = mean(total_revenue, na.rm = TRUE),
.groups = "drop"
)
rev_by_pop_wide <- rev_by_pop %>%
select(pop_bracket, lgbtq_label, N, median_rev) %>%
pivot_wider(names_from = lgbtq_label,
values_from = c(N, median_rev),
names_sep = "_") %>%
mutate(
ratio = round(`median_rev_LGBTQ+` / `median_rev_Non-LGBTQ+`, 2)
)
rev_by_pop_wide %>%
transmute(
`Pop. Bracket` = pop_bracket,
`N LGBTQ+` = format_n(`N_LGBTQ+`),
`Median LGBTQ+` = format_brl(`median_rev_LGBTQ+`),
`N Non-LGBTQ+` = format_n(`N_Non-LGBTQ+`),
`Median Non-LGBTQ+` = format_brl(`median_rev_Non-LGBTQ+`),
`Ratio` = ratio
) %>%
kable(align = c("l", rep("r", 5)))| Pop. Bracket | N LGBTQ+ | Median LGBTQ+ | N Non-LGBTQ+ | Median Non-LGBTQ+ | Ratio |
|---|---|---|---|---|---|
| < 10K | 314 | R$ 1,429 | 93,165 | R$ 1,800 | 0.79 |
| 10K-50K | 867 | R$ 3,000 | 158,638 | R$ 2,368 | 1.27 |
| 50K-200K | 723 | R$ 5,000 | 75,823 | R$ 3,708 | 1.35 |
| 200K-500K | 367 | R$12,851 | 27,452 | R$ 7,498 | 1.71 |
| 500K+ | 434 | R$35,356 | 19,920 | R$15,517 | 2.28 |
ggplot(rev_by_pop, aes(x = pop_bracket, y = median_rev, fill = lgbtq_label)) +
geom_col(position = "dodge", alpha = 0.85) +
scale_y_continuous(labels = scales::label_dollar(prefix = "R$", big.mark = ",")) +
scale_fill_manual(values = pal_lgbtq) +
labs(
x = "Municipality Population",
y = "Median Revenue",
fill = NULL,
title = "Revenue by Municipality Size and LGBTQ+ Status",
subtitle = "Among candidates with non-zero revenue"
)
save_figure(last_plot(), "05_revenue_by_pop_bracket")
LGBTQ+ candidates disproportionately run in larger municipalities where campaigns are more expensive. The aggregate revenue comparison may therefore obscure the within-market fundraising gap. Compare the overall median ratio with the bracket-specific ratios above.
This section pools across position types because the analysis controls for municipality size directly. The comparison examines whether LGBTQ+ candidates raise more or less than non-LGBTQ+ counterparts within the same population bracket, making position disaggregation less critical. Mayoral campaigns operate at a fundamentally different revenue scale; the position-specific tabs in earlier sections capture this difference.
The within-ideology revenue ratio compares LGBTQ+ and non-LGBTQ+ candidates within the same ideological bloc (Left, Center, Right). The ratio is computed as LGBTQ+ mean (or median) revenue divided by non-LGBTQ+ mean (or median) revenue for each bloc. A ratio below 1 means LGBTQ+ candidates raise less than their ideological peers; a ratio above 1 means they raise more.
within_ideo <- df %>%
filter(!is.na(ideology_category)) %>%
mutate(group = if_else(lgbtq_candidate, "LGBTQ+", "Non-LGBTQ+")) %>%
group_by(ideology_category, group) %>%
summarise(
N = n(),
mean_rev = mean(total_revenue, na.rm = TRUE),
median_rev = median(total_revenue, na.rm = TRUE),
.groups = "drop"
)
within_ideo_wide <- within_ideo %>%
select(ideology_category, group, N, mean_rev, median_rev) %>%
pivot_wider(
names_from = group,
values_from = c(N, mean_rev, median_rev),
names_glue = "{group}_{.value}"
) %>%
mutate(
mean_ratio = if_else(`Non-LGBTQ+_mean_rev` > 0,
round(`LGBTQ+_mean_rev` / `Non-LGBTQ+_mean_rev`, 2), NA_real_),
median_ratio = if_else(`Non-LGBTQ+_median_rev` > 0,
round(`LGBTQ+_median_rev` / `Non-LGBTQ+_median_rev`, 2), NA_real_)
) %>%
mutate(ideology_category = factor(ideology_category, levels = ideology_levels)) %>%
arrange(ideology_category)
within_ideo_wide %>%
transmute(
Ideology = ideology_category,
`N LGBTQ+` = format_n(`LGBTQ+_N`),
`N Non-LGBTQ+` = format_n(`Non-LGBTQ+_N`),
`Mean Rev. LGBTQ+` = format_brl(`LGBTQ+_mean_rev`),
`Mean Rev. Non-LGBTQ+` = format_brl(`Non-LGBTQ+_mean_rev`),
`Mean Ratio (L/NL)` = if_else(is.na(mean_ratio), "---", as.character(mean_ratio)),
`Median Rev. LGBTQ+` = format_brl(`LGBTQ+_median_rev`),
`Median Rev. Non-LGBTQ+` = format_brl(`Non-LGBTQ+_median_rev`),
`Median Ratio (L/NL)` = if_else(is.na(median_ratio), "---", as.character(median_ratio))
) %>%
kable(align = c("l", "r", "r", "r", "r", "r", "r", "r", "r"))| Ideology | N LGBTQ+ | N Non-LGBTQ+ | Mean Rev. LGBTQ+ | Mean Rev. Non-LGBTQ+ | Mean Ratio (L/NL) | Median Rev. LGBTQ+ | Median Rev. Non-LGBTQ+ | Median Ratio (L/NL) |
|---|---|---|---|---|---|---|---|---|
| Left | 1,256 | 59,474 | R$42,155 | R$21,966 | 1.92 | R$9,108 | R$2,437 | 3.74 |
| Center | 991 | 165,755 | R$13,336 | R$14,014 | 0.95 | R$2,000 | R$1,455 | 1.37 |
| Right | 887 | 235,238 | R$10,231 | R$16,913 | 0.6 | R$1,841 | R$1,800 | 1.02 |
plot_data <- within_ideo_wide %>%
filter(!is.na(mean_ratio) | !is.na(median_ratio)) %>%
select(ideology_category, `Mean ratio` = mean_ratio, `Median ratio` = median_ratio) %>%
pivot_longer(-ideology_category, names_to = "metric", values_to = "ratio") %>%
filter(!is.na(ratio))
ggplot(plot_data, aes(x = ideology_category, y = ratio,
color = metric, shape = metric)) +
geom_hline(yintercept = 1, linetype = "dashed", color = "gray50") +
geom_point(size = 4, position = position_dodge(width = 0.4)) +
coord_flip() +
scale_color_manual(values = c("Mean ratio" = "#E74C3C", "Median ratio" = "#3498DB"),
name = NULL) +
scale_shape_manual(values = c("Mean ratio" = 16, "Median ratio" = 17),
name = NULL) +
labs(
x = NULL,
y = "Revenue Ratio (LGBTQ+ / Non-LGBTQ+)",
title = "Within-Ideology Revenue Ratios",
subtitle = "Ratio < 1 means LGBTQ+ candidates raise less than ideological peers",
caption = "Dashed line = parity (ratio = 1). Only candidates with assigned ideology included."
)
save_figure(last_plot(), "05_within_ideology_ratio")
n_ideo_below1_mean <- sum(within_ideo_wide$mean_ratio < 1, na.rm = TRUE)
n_ideo_above1_mean <- sum(within_ideo_wide$mean_ratio > 1, na.rm = TRUE)
n_ideo_below1_median <- sum(within_ideo_wide$median_ratio < 1, na.rm = TRUE)
n_ideo_total <- nrow(within_ideo_wide)Of the 3 ideological blocs, 2 show a mean revenue ratio below 1 (LGBTQ+ candidates raising less than ideological peers) and 1 show a mean ratio above 1. At the median, 0 of 3 blocs have ratios below 1.
The within-ideology revenue comparison pools across position types because it already conditions on a key confounder (ideology). Position-specific results for the LGBTQ+ vs. non-LGBTQ+ revenue comparison are available in the tabbed section above.
The TSE classifies each transaction’s DS_NATUREZA_RECEITA as either “FINANCEIRO” (financial) or a non-financial category. Financial contributions are direct monetary transfers: cash, PIX, bank deposits, and electronic transfers. In-kind contributions (recursos estimaveis) are non-monetary — goods, services, or volunteer labor that the campaign assigns a monetary value to for accounting purposes. Examples include donated office space, printing services, vehicles for campaign use, or professional labor contributed without charge. The variables pct_financial and pct_inkind express each type’s share of a candidate’s total revenue on a 0–100 scale.
render_financial_inkind <- function(data, tab_name) {
# --- Financial vs In-Kind comparison table ---
cat("### Financial vs In-Kind Revenue\n\n")
data %>%
mutate(group = if_else(lgbtq_candidate, "LGBTQ+", "Non-LGBTQ+")) %>%
group_by(group) %>%
summarise(
N = format_n(n()),
`Total Financial` = format_brl(sum(financial_amt, na.rm = TRUE)),
`Total In-Kind` = format_brl(sum(inkind_amt, na.rm = TRUE)),
`Mean % Financial` = format_pct100(mean(pct_financial, na.rm = TRUE)),
`Mean % In-Kind` = format_pct100(mean(pct_inkind, na.rm = TRUE)),
`Median % Financial` = format_pct100(median(pct_financial, na.rm = TRUE)),
`Median % In-Kind` = format_pct100(median(pct_inkind, na.rm = TRUE)),
.groups = "drop"
) %>%
rename(Group = group) %>%
cat_kable(align = c("l", "r", "r", "r", "r", "r", "r", "r"))
# --- Density plot of financial share ---
cat("### Financial Share Distribution\n\n")
pos_rev <- data %>% filter(total_revenue > 0)
if (nrow(pos_rev) > 0) {
p_density <- pos_rev %>%
mutate(group = if_else(lgbtq_candidate, "LGBTQ+", "Non-LGBTQ+")) %>%
ggplot(aes(x = pct_financial / 100, fill = group, color = group)) +
geom_density(alpha = 0.3, linewidth = 0.8) +
scale_fill_manual(values = pal_lgbtq, name = NULL) +
scale_color_manual(values = pal_lgbtq, name = NULL) +
scale_x_continuous(labels = percent, limits = c(0, 1)) +
labs(
x = "Financial Revenue as % of Total Revenue",
y = "Density",
title = paste0("Share of Financial (vs In-Kind) Revenue (", tab_name, ")"),
subtitle = "Among candidates with positive revenue"
)
cat_plot(p_density, paste0("05-financial-inkind-density-", pos_suffix(tab_name)))
}
# --- Stacked bar by identity category ---
cat("### Financial vs In-Kind by Identity\n\n")
p_identity <- data %>%
filter(total_revenue > 0) %>%
mutate(
plot_category = if_else(lgbtq_candidate, as.character(lgbt_category), "Non-LGBTQ+"),
plot_category = factor(plot_category,
levels = c("Non-LGBTQ+", "Gay", "Lesbian", "Bisexual+",
"Trans", "Asexual"))
) %>%
filter(!is.na(plot_category)) %>%
group_by(plot_category) %>%
summarise(
Financial = sum(financial_amt, na.rm = TRUE),
`In-Kind` = sum(inkind_amt, na.rm = TRUE),
.groups = "drop"
) %>%
pivot_longer(-plot_category, names_to = "type", values_to = "total") %>%
group_by(plot_category) %>%
mutate(pct = total / sum(total)) %>%
ungroup() %>%
ggplot(aes(x = plot_category, y = pct, fill = type)) +
geom_col(alpha = 0.9) +
geom_text(aes(label = format_pct(pct)),
position = position_stack(vjust = 0.5), size = 3.5, color = "white",
fontface = "bold") +
scale_fill_manual(values = c("Financial" = "#3498DB", "In-Kind" = "#E67E22"),
name = "Revenue Type") +
scale_y_continuous(labels = percent) +
labs(
x = NULL,
y = "Share of Revenue",
title = paste0("Financial vs In-Kind Revenue Composition (", tab_name, ")"),
subtitle = "By identity category (among candidates with positive revenue)"
) +
theme(axis.text.x = element_text(angle = 30, hjust = 1))
cat_plot(p_identity, paste0("05-financial-inkind-identity-", pos_suffix(tab_name)))
cat("::: {.callout-note}\n")
cat("## Financial vs. In-Kind\n")
cat("Financial contributions are direct monetary transfers (cash, PIX, bank transfers). ")
cat("In-kind contributions are non-monetary goods and services assigned a monetary value ")
cat("for accounting purposes. The table and density plot above compare these two categories ")
cat("across LGBTQ+ and non-LGBTQ+ candidates.\n")
cat(":::\n\n")
}
render_position_tabset(render_financial_inkind, df)| Group | N | Total Financial | Total In-Kind | Mean % Financial | Mean % In-Kind | Median % Financial | Median % In-Kind |
|---|---|---|---|---|---|---|---|
| LGBTQ+ | 3,043 | R$50,691,864 | R$6,126,029 | 61.3% | 26.5% | 82.4% | 5.5% |
| Non-LGBTQ+ | 428,962 | R$2,593,493,902 | R$524,080,400 | 49.9% | 34.0% | 58.2% | 11.5% |
Financial contributions are direct monetary transfers (cash, PIX, bank transfers). In-kind contributions are non-monetary goods and services assigned a monetary value for accounting purposes. The table and density plot above compare these two categories across LGBTQ+ and non-LGBTQ+ candidates.
| Group | N | Total Financial | Total In-Kind | Mean % Financial | Mean % In-Kind | Median % Financial | Median % In-Kind |
|---|---|---|---|---|---|---|---|
| LGBTQ+ | 91 | R$18,011,658 | R$407,030 | 36.3% | 5.4% | 0.0% | 0.0% |
| Non-LGBTQ+ | 31,505 | R$4,375,905,490 | R$114,323,831 | 45.9% | 2.5% | 0.0% | 0.0% |
Financial contributions are direct monetary transfers (cash, PIX, bank transfers). In-kind contributions are non-monetary goods and services assigned a monetary value for accounting purposes. The table and density plot above compare these two categories across LGBTQ+ and non-LGBTQ+ candidates.
This tab pools city councilors (proportional representation) and mayors/vice-mayors (plurality). Position-specific results in the other tabs may be more informative.
| Group | N | Total Financial | Total In-Kind | Mean % Financial | Mean % In-Kind | Median % Financial | Median % In-Kind |
|---|---|---|---|---|---|---|---|
| LGBTQ+ | 3,134 | R$68,703,522 | R$6,533,059 | 60.6% | 25.9% | 81.9% | 4.9% |
| Non-LGBTQ+ | 460,467 | R$6,969,399,393 | R$638,404,231 | 49.6% | 31.9% | 57.6% | 7.7% |
Financial contributions are direct monetary transfers (cash, PIX, bank transfers). In-kind contributions are non-monetary goods and services assigned a monetary value for accounting purposes. The table and density plot above compare these two categories across LGBTQ+ and non-LGBTQ+ candidates.
The in-kind contribution type analysis below pools across city councilors and mayors/vice-mayors. This section examines transaction-level data (individual contribution records) rather than candidate-level aggregates, and the types of in-kind support received do not vary systematically by position type.
The TSE records the specific type of each in-kind contribution in the field DS_NATUREZA_RECURSO_ESTIMAVEL. This variable is only populated for non-financial transactions and describes the nature of the donated good or service (e.g., advertising materials, vehicles, professional services). This section examines which types of in-kind support are most common and whether the composition differs between LGBTQ+ and non-LGBTQ+ candidates.
# Filter transaction-level data to in-kind contributions
inkind_trans <- receitas %>%
filter(DS_NATUREZA_RECEITA != "FINANCEIRO") %>%
filter(!is.na(DS_NATUREZA_RECURSO_ESTIMAVEL),
DS_NATUREZA_RECURSO_ESTIMAVEL != "#NULO#",
DS_NATUREZA_RECURSO_ESTIMAVEL != "#NULO")
# Join with df to get LGBTQ+ flag (candidate_id in df == SQ_CANDIDATO in transactions)
inkind_trans <- inkind_trans %>%
inner_join(
df %>% select(candidate_id, lgbtq_candidate),
by = c("SQ_CANDIDATO" = "candidate_id")
) %>%
mutate(group = if_else(lgbtq_candidate, "LGBTQ+", "Non-LGBTQ+"))
# Translate and group Portuguese categories into English labels
# Top categories get their own label; rare categories are grouped
inkind_trans <- inkind_trans %>%
mutate(
inkind_en = case_when(
str_detect(DS_NATUREZA_RECURSO_ESTIMAVEL, "materiais impressos") ~ "Printed materials",
str_detect(DS_NATUREZA_RECURSO_ESTIMAVEL, "adesivos") ~ "Stickers/decals",
str_detect(DS_NATUREZA_RECURSO_ESTIMAVEL, "veículos") ~ "Vehicle use",
str_detect(DS_NATUREZA_RECURSO_ESTIMAVEL, "contábeis") ~ "Accounting services",
str_detect(DS_NATUREZA_RECURSO_ESTIMAVEL, "advocatícios") ~ "Legal services",
str_detect(DS_NATUREZA_RECURSO_ESTIMAVEL, "militância") ~ "Grassroots mobilization",
str_detect(DS_NATUREZA_RECURSO_ESTIMAVEL, "rádio|televisão|vídeo") ~ "Media production",
str_detect(DS_NATUREZA_RECURSO_ESTIMAVEL, "terceiros") ~ "Third-party services",
str_detect(DS_NATUREZA_RECURSO_ESTIMAVEL, "pessoal") ~ "Personnel costs",
str_detect(DS_NATUREZA_RECURSO_ESTIMAVEL, "jingles|vinhetas") ~ "Jingles/slogans",
str_detect(DS_NATUREZA_RECURSO_ESTIMAVEL, "bens imóveis") ~ "Property rental",
TRUE ~ "Other"
)
)inkind_freq <- inkind_trans %>%
group_by(inkind_en) %>%
summarise(
n = n(),
total_amt = sum(VR_RECEITA, na.rm = TRUE),
mean_amt = mean(VR_RECEITA, na.rm = TRUE),
.groups = "drop"
) %>%
arrange(desc(n)) %>%
mutate(
pct = n / sum(n),
cum_pct = cumsum(pct),
pct_amt = total_amt / sum(total_amt)
)
n_inkind_types_raw <- inkind_trans %>%
pull(DS_NATUREZA_RECURSO_ESTIMAVEL) %>%
n_distinct()
inkind_freq %>%
mutate(
`N Transactions` = format_n(n),
`% Trans.` = format_pct(pct),
`Total (R$)` = format_brl(total_amt),
`% Value` = format_pct(pct_amt),
`Mean (R$)` = format_brl(mean_amt)
) %>%
select(
`In-Kind Type` = inkind_en,
`N Transactions`,
`% Trans.`,
`Total (R$)`,
`% Value`,
`Mean (R$)`
) %>%
kable(align = c("l", "r", "r", "r", "r", "r"))| In-Kind Type | N Transactions | % Trans. | Total (R$) | % Value | Mean (R$) |
|---|---|---|---|---|---|
| Printed materials | 402,391 | 37.7% | R$127,619,133 | 19.8% | R$ 317 |
| Stickers/decals | 225,791 | 21.1% | R$ 47,559,910 | 7.4% | R$ 211 |
| Vehicle use | 87,371 | 8.2% | R$172,307,206 | 26.7% | R$1,972 |
| Accounting services | 74,987 | 7.0% | R$ 40,885,990 | 6.3% | R$ 545 |
| Legal services | 66,226 | 6.2% | R$ 40,287,414 | 6.3% | R$ 608 |
| Grassroots mobilization | 51,487 | 4.8% | R$ 45,442,873 | 7.1% | R$ 883 |
| Other | 37,193 | 3.5% | R$ 32,443,757 | 5.0% | R$ 872 |
| Third-party services | 37,081 | 3.5% | R$ 42,643,395 | 6.6% | R$1,150 |
| Media production | 32,889 | 3.1% | R$ 40,478,701 | 6.3% | R$1,231 |
| Personnel costs | 25,582 | 2.4% | R$ 30,296,495 | 4.7% | R$1,184 |
| Jingles/slogans | 16,598 | 1.6% | R$ 10,329,620 | 1.6% | R$ 622 |
| Property rental | 11,013 | 1.0% | R$ 14,227,509 | 2.2% | R$1,292 |
# Top in-kind type
top_inkind_type <- inkind_freq$inkind_en[1]
top_inkind_pct <- inkind_freq$pct[1]
second_inkind_type <- inkind_freq$inkind_en[2]
second_inkind_pct <- inkind_freq$pct[2]
top6_cum <- inkind_freq$cum_pct[min(6, nrow(inkind_freq))]The TSE records 39 distinct in-kind categories; we group these into 12 English-translated types for readability. The two largest — Printed materials (37.7%) and Stickers/decals (21.1%) — together account for nearly 58.8% of all in-kind transactions. The top six types cover 85.0%.
# Join identity to in-kind transactions
inkind_identity <- inkind_trans %>%
inner_join(
df %>% select(candidate_id, lgbt_category),
by = c("SQ_CANDIDATO" = "candidate_id")
)
inkind_by_group <- inkind_identity %>%
mutate(
display_group = if_else(lgbtq_candidate, as.character(lgbt_category), "Non-LGBTQ+")
) %>%
group_by(display_group) %>%
summarise(
`N Transactions` = n(),
`Total (R$)` = format_brl(sum(VR_RECEITA, na.rm = TRUE)),
`Mean (R$)` = format_brl(mean(VR_RECEITA, na.rm = TRUE)),
`Median (R$)` = format_brl(median(VR_RECEITA, na.rm = TRUE)),
.groups = "drop"
)
inkind_by_group %>%
rename(Group = display_group) %>%
kable(align = c("l", "r", "r", "r", "r"),
format.args = list(big.mark = ","))| Group | N Transactions | Total (R$) | Mean (R$) | Median (R$) |
|---|---|---|---|---|
| Asexual | 308 | R$169,730 | R$551 | R$250 |
| Bisexual+ | 1,316 | R$1,719,589 | R$1,307 | R$446 |
| Gay | 2,515 | R$2,071,531 | R$824 | R$300 |
| Lesbian | 1,682 | R$1,458,824 | R$867 | R$300 |
| Non-LGBTQ+ | 1,065,400 | R$651,240,677 | R$611 | R$210 |
| Trans | 1,446 | R$1,125,658 | R$778 | R$300 |
# Compute shares by group for the top types (excluding "Other")
top_labels <- inkind_freq %>%
filter(inkind_en != "Other") %>%
head(8) %>%
pull(inkind_en)
inkind_by_group <- inkind_trans %>%
filter(inkind_en %in% top_labels) %>%
count(group, inkind_en) %>%
group_by(group) %>%
mutate(pct = n / sum(n)) %>%
ungroup() %>%
mutate(inkind_en = factor(inkind_en, levels = rev(top_labels)))
ggplot(inkind_by_group, aes(x = pct, y = inkind_en, fill = group)) +
geom_col(position = "dodge", alpha = 0.85, width = 0.7) +
geom_text(aes(label = format_pct(pct)),
position = position_dodge(width = 0.7), hjust = -0.1,
size = 3) +
scale_fill_manual(values = pal_lgbtq, name = NULL) +
scale_x_continuous(labels = percent, expand = expansion(mult = c(0, 0.15))) +
labs(
x = "Share of In-Kind Transactions (within group)",
y = NULL,
title = "In-Kind Contribution Types: LGBTQ+ vs Non-LGBTQ+",
subtitle = "Top 8 types shown; shares computed within each group's in-kind total",
caption = "Categories translated from TSE's DS_NATUREZA_RECURSO_ESTIMAVEL field."
)
save_figure(last_plot(), "05_inkind_types_comparison", height = 7)
# Compare top type across groups
lgbtq_top <- inkind_trans %>%
filter(group == "LGBTQ+") %>%
count(inkind_en, sort = TRUE) %>%
mutate(pct = n / sum(n)) %>%
slice_max(pct, n = 1)
nonlgbtq_top <- inkind_trans %>%
filter(group == "Non-LGBTQ+") %>%
count(inkind_en, sort = TRUE) %>%
mutate(pct = n / sum(n)) %>%
slice_max(pct, n = 1)
# Find biggest difference between groups
inkind_diff <- inkind_trans %>%
filter(inkind_en %in% top_labels) %>%
count(group, inkind_en) %>%
group_by(group) %>%
mutate(pct = n / sum(n)) %>%
ungroup() %>%
select(group, inkind_en, pct) %>%
pivot_wider(names_from = group, values_from = pct, values_fill = 0) %>%
mutate(diff = `LGBTQ+` - `Non-LGBTQ+`) %>%
slice_max(abs(diff), n = 1)Among LGBTQ+ candidates, the most common in-kind type is Printed materials (37.6%), while for non-LGBTQ+ candidates it is Printed materials (37.7%). The largest compositional difference is in Media production, where the LGBTQ+ share is 3.3 percentage points higher than the non-LGBTQ+ share.
# Funding composition differences
party_diff_direction <- if (lgbtq_pct_party > nonlgbtq_pct_party) "higher" else if (lgbtq_pct_party < nonlgbtq_pct_party) "lower" else "similar"
indiv_diff_direction <- if (lgbtq_pct_individual > nonlgbtq_pct_individual) "higher" else if (lgbtq_pct_individual < nonlgbtq_pct_individual) "lower" else "similar"
# Identity categories
n_identity_cats <- df %>%
filter(lgbtq_candidate, lgbt_category != "Other LGBTQ+") %>%
pull(lgbt_category) %>%
n_distinct()
# Highest and lowest median revenue among identity categories
identity_median_stats <- df %>%
filter(lgbtq_candidate, lgbt_category != "Other LGBTQ+") %>%
group_by(lgbt_category) %>%
summarise(median_rev = median(total_revenue), .groups = "drop")
top_median_cat <- identity_median_stats %>% slice_max(median_rev, n = 1) %>% pull(lgbt_category)
top_median_val <- identity_median_stats %>% slice_max(median_rev, n = 1) %>% pull(median_rev)
bottom_median_cat <- identity_median_stats %>% slice_min(median_rev, n = 1) %>% pull(lgbt_category)
bottom_median_val <- identity_median_stats %>% slice_min(median_rev, n = 1) %>% pull(median_rev)This chapter documents the financial landscape of LGBTQ+ candidacies in Brazil’s 2024 municipal elections. The key patterns are:
Revenue scale: Campaign finance in municipal elections spans several orders of magnitude, from 85,653 candidates (18.5%) with zero reported revenue to a maximum of R$81,590,273. The median campaign raised R$1,752.
LGBTQ+ revenue comparison: LGBTQ+ candidates have a median revenue of R$3,705, which is higher than the non-LGBTQ+ median of R$1,746 (ratio: 2.12). At the mean, LGBTQ+ candidates raise R$24,007 versus R$16,522 (ratio: 1.45).
Funding composition: LGBTQ+ candidates have a higher average party funding share (47.6% vs 27.2%) and a lower average individual donor share (12.6% vs 17.3%) compared to non-LGBTQ+ candidates.
Within-ideology patterns: Of 3 ideological blocs examined, 2 show a mean revenue ratio below 1 (LGBTQ+ candidates raising less than ideological peers) and 1 show a ratio above 1.
Identity-specific patterns: Across 5 identity categories, Bisexual+ candidates have the highest median revenue (R$6,460) and Asexual candidates the lowest (R$1,330). Bisexual+ candidates rely most heavily on party funding (58.9% of revenue on average).
The next chapters examine intersectional patterns and geographic variation.