1 Overview

This document estimates the deposit provision cost \(c\) following the DSSW (2026, forthcoming) hedonic approach. The key idea: net non-interest expense includes both deposit-servicing costs and non-deposit overhead. A hedonic regression isolates the marginal cost of servicing each dollar of deposits.

Methodology:

\[ \frac{\text{Net Non-Interest Expense}_{i,t}}{\text{Total Assets}_{i,t}} = \alpha_t + \sum_k \beta_k \frac{\text{Deposit}_k}{\text{TA}}_{i,t} + \gamma' \mathbf{X}_{i,t} + \varepsilon_{i,t} \]

where \(k\) indexes deposit categories (insured, uninsured, time) and \(\mathbf{X}\) includes balance-sheet controls (loans, equity, FHLB, etc.).

The estimated \(\hat{\beta}_k\) coefficients represent the marginal non-interest cost per dollar of each deposit type. The bank-level deposit cost is then:

\[ c_i = \sum_k \hat{\beta}_k \frac{\text{Deposit}_{k,i}}{\text{Total Deposits}_i} \]

Output: 01_data/processed/dssw_deposit_costs.csv

2 Setup

rm(list = ls())
library(data.table); library(dplyr); library(tidyr); library(stringr)
library(lubridate); library(purrr); library(tibble)
library(fixest); library(modelsummary)
library(knitr); library(kableExtra)
library(ggplot2); library(scales)
library(readr)
cat("All packages loaded.\n")
## All packages loaded.

2.1 Paths

BASE_PATH   <- "C:/Users/mferdo2/OneDrive - Louisiana State University/Finance_PhD/DW_Stigma_paper/Liquidity_project_2025"
DATA_RAW    <- file.path(BASE_PATH, "01_data/raw")
DATA_PROC   <- file.path(BASE_PATH, "01_data/processed")
OUTPUT_PATH <- file.path(BASE_PATH, "01_data/processed")
TABLE_PATH  <- file.path(OUTPUT_PATH, "tables")
FIG_PATH    <- file.path(OUTPUT_PATH, "figures")
for (path in c(TABLE_PATH, FIG_PATH)) if (!dir.exists(path)) dir.create(path, recursive = TRUE)

3 Data Loading

We load the full clean call report (which contains income statement items from Schedule RI) — not the final processed panel which drops these columns.

# ==============================================================================
# Load clean call report with all RIAD (income statement) variables
# This file has: noninterest_income, noninterest_expense, 
#   total_interest_on_deposits_calc, and component-level deposit interest
# ==============================================================================

call_full <- read_csv(file.path(DATA_PROC, "clean_call_report_with_mtm_2021Q4_2024Q4.csv"),
                      show_col_types = FALSE) %>%
  mutate(idrssd = as.character(idrssd),
         period = paste0(year, "Q", quarter))
## Warning: One or more parsing issues, call `problems()` on your data frame for details,
## e.g.:
##   dat <- vroom(...)
##   problems(dat)
# Load GSIB flags to exclude
gsib_panel <- read_csv(file.path(DATA_PROC, "final_call_gsib.csv"), show_col_types = FALSE) %>%
  mutate(idrssd = as.character(idrssd)) %>%
  select(idrssd, period, gsib, failed_bank, size_bin) %>%
  distinct()

cat("Call report rows:", nrow(call_full), "\n")
## Call report rows: 75989
cat("Available periods:", paste(sort(unique(call_full$period)), collapse = ", "), "\n")
## Available periods: 2021Q1, 2021Q2, 2021Q3, 2021Q4, 2022Q1, 2022Q2, 2022Q3, 2022Q4, 2023Q1, 2023Q2, 2023Q3, 2023Q4, 2024Q1, 2024Q2, 2024Q3, 2024Q4

3.1 Check Available Variables

# Verify income/expense columns exist
required_vars <- c("noninterest_income", "noninterest_expense",
                    "total_interest_on_deposits_calc",
                    "insured_deposit", "uninsured_deposit",
                    "total_deposit", "total_asset", "total_loan",
                    "total_bank_equity", "fhlb_adv")

available <- required_vars[required_vars %in% names(call_full)]
missing   <- required_vars[!required_vars %in% names(call_full)]

cat("Available:", paste(available, collapse = ", "), "\n")
## Available: noninterest_income, noninterest_expense, total_interest_on_deposits_calc, insured_deposit, uninsured_deposit, total_deposit, total_asset, total_loan, total_bank_equity, fhlb_adv
if (length(missing) > 0) cat("MISSING:", paste(missing, collapse = ", "), "\n")

4 Variable Construction

# ==============================================================================
# Construct hedonic regression variables
# All scaled by total assets for cross-sectional comparability
# ==============================================================================

df <- call_full %>%
 
  mutate(
    # ── LHS: Net non-interest expense / TA ──
    net_nie_to_ta = (noninterest_expense - noninterest_income) / total_asset,

    # ── Deposit subcategories / TA ──
    # Insured deposits (predominantly zero-maturity: transaction + savings)
    insured_dep_to_ta   = insured_deposit / total_asset,
    # Uninsured deposits (predominantly zero-maturity demand deposits)
    uninsured_dep_to_ta = uninsured_deposit / total_asset,

    # ── Controls / TA ──
    loans_to_ta     = ifelse(!is.na(total_loan), total_loan / total_asset, NA_real_),
    equity_to_ta    = ifelse(!is.na(total_bank_equity), total_bank_equity / total_asset, NA_real_),
    fhlb_to_ta      = ifelse(!is.na(fhlb_adv), fhlb_adv / total_asset, 0),
    other_borr_to_ta = ifelse(!is.na(other_borrowed_money),
                              other_borrowed_money / total_asset, 0),
    fed_funds_to_ta  = ifelse(!is.na(fed_fund_purchase),
                              fed_fund_purchase / total_asset, 0),
    repo_to_ta       = ifelse(!is.na(repo), repo / total_asset, 0),

    # ── Size quartiles (for interactions) ──
    ln_ta = log(total_asset),

    # ── Interest cost of deposits (for comparison) ──
    int_cost_dep = total_interest_on_deposits_calc / total_deposit,

    # ── Total deposit cost (interest + net non-interest) ──
    total_dep_to_ta = total_deposit / total_asset
  )

# Size quartiles within each period
df <- df %>%
  group_by(period) %>%
  mutate(size_quartile = ntile(total_asset, 4)) %>%
  ungroup()

cat("Analysis sample:", nrow(df), "bank-quarters\n")
## Analysis sample: 75989 bank-quarters
cat("Periods:", n_distinct(df$period), "| Banks:", n_distinct(df$idrssd), "\n")
## Periods: 16 | Banks: 5074

4.1 Summary Statistics

summ_vars <- c("net_nie_to_ta", "insured_dep_to_ta", "uninsured_dep_to_ta",
               "loans_to_ta", "equity_to_ta", "fhlb_to_ta", "int_cost_dep")

df_summ <- df %>%
  select(all_of(summ_vars)) %>%
  pivot_longer(everything(), names_to = "Variable", values_to = "Value") %>%
  group_by(Variable) %>%
  summarise(N = sum(!is.na(Value)),
            Mean = round(mean(Value, na.rm = TRUE), 5),
            SD = round(sd(Value, na.rm = TRUE), 5),
            P25 = round(quantile(Value, 0.25, na.rm = TRUE), 5),
            Median = round(median(Value, na.rm = TRUE), 5),
            P75 = round(quantile(Value, 0.75, na.rm = TRUE), 5),
            .groups = "drop")

kbl(df_summ, format = "html", caption = "Summary: Hedonic Regression Variables") %>%
  kable_styling(bootstrap_options = c("striped","hover","condensed"), full_width = FALSE)
Summary: Hedonic Regression Variables
Variable N Mean SD P25 Median P75
equity_to_ta 75987 0.11673 0.10744 0.08097 0.09682 0.11834
fhlb_to_ta 75987 0.02480 0.04182 0.00000 0.00107 0.03538
insured_dep_to_ta 75987 0.61897 0.15124 0.55856 0.64593 0.71643
int_cost_dep 75164 Inf NaN 0.00134 0.00322 0.00832
loans_to_ta 75987 0.59714 0.18289 0.49741 0.62437 0.73226
net_nie_to_ta 75987 0.00904 0.06216 0.00571 0.01053 0.01598
uninsured_dep_to_ta 75987 0.21927 0.12190 0.13664 0.20362 0.28242

5 Hedonic Deposit Cost Regression

5.1 Main Specification (DSSW-style)

# ==============================================================================
# DSSW Hedonic Regression (Eq. 67)
# 
# LHS: Net Non-Interest Expense / Total Assets
# RHS: Deposit subcategories / TA + Controls + Period FE
#
# The βs on deposit types = marginal non-interest cost per dollar of deposits
# ==============================================================================

# Winsorize at 2.5%/97.5% to reduce outlier influence
winsorize <- function(x, probs = c(0.025, 0.975)) {
  if (all(is.na(x))) return(x)
  q <- quantile(x, probs = probs, na.rm = TRUE, names = FALSE)
  pmax(pmin(x, q[2]), q[1])
}

df_reg <- df %>%
  filter(!is.na(net_nie_to_ta), !is.na(insured_dep_to_ta),
         !is.na(uninsured_dep_to_ta), !is.na(loans_to_ta),
         !is.na(equity_to_ta)) %>%
  mutate(across(c(net_nie_to_ta, insured_dep_to_ta, uninsured_dep_to_ta,
                  loans_to_ta, equity_to_ta, fhlb_to_ta,
                  other_borr_to_ta, fed_funds_to_ta, repo_to_ta),
                winsorize))

# ── Specification 1: Basic (deposit types + controls) ──
reg1 <- feols(net_nie_to_ta ~ insured_dep_to_ta + uninsured_dep_to_ta +
                loans_to_ta + equity_to_ta + fhlb_to_ta + ln_ta | period,
              data = df_reg, cluster = ~idrssd)

# ── Specification 2: Full controls ──
reg2 <- feols(net_nie_to_ta ~ insured_dep_to_ta + uninsured_dep_to_ta +
                loans_to_ta + equity_to_ta + fhlb_to_ta +
                other_borr_to_ta + fed_funds_to_ta + repo_to_ta + ln_ta | period,
              data = df_reg, cluster = ~idrssd)

# ── Specification 3: Size-quartile interactions ──
reg3 <- feols(net_nie_to_ta ~ insured_dep_to_ta * factor(size_quartile) +
                uninsured_dep_to_ta +
                loans_to_ta + equity_to_ta + fhlb_to_ta +
                other_borr_to_ta + ln_ta | period,
              data = df_reg, cluster = ~idrssd)

# ── Specification 4: Cross-sectional (2022Q4 only) ──
df_2022q4 <- df_reg %>% filter(period == "2022Q4")
reg4 <- feols(net_nie_to_ta ~ insured_dep_to_ta + uninsured_dep_to_ta +
                loans_to_ta + equity_to_ta + fhlb_to_ta +
                other_borr_to_ta + ln_ta,
              data = df_2022q4, vcov = "hetero")

models <- list(
  "Basic + FE"         = reg1,
  "Full Controls + FE" = reg2,
  "Size × Insured"     = reg3,
  "2022Q4 Only"        = reg4
)

msummary(models, stars = c("*" = .10, "**" = .05, "***" = .01),
         gof_omit = "AIC|BIC|Log|RMSE",
         coef_rename = c(
           "insured_dep_to_ta" = "Insured Deposits / TA",
           "uninsured_dep_to_ta" = "Uninsured Deposits / TA",
           "loans_to_ta" = "Loans / TA",
           "equity_to_ta" = "Equity / TA",
           "fhlb_to_ta" = "FHLB / TA",
           "other_borr_to_ta" = "Other Borrowing / TA",
           "fed_funds_to_ta" = "Fed Funds Purchased / TA",
           "repo_to_ta" = "Repo / TA",
           "ln_ta" = "Log(Total Assets)"),
         title = "Hedonic Deposit Cost Regression (DSSW Approach)")
Hedonic Deposit Cost Regression (DSSW Approach)
Basic + FE Full Controls + FE Size × Insured 2022Q4 Only
* p < 0.1, ** p < 0.05, *** p < 0.01
Insured Deposits / TA 0.013*** 0.014*** 0.014*** 0.021***
(0.001) (0.001) (0.001) (0.002)
Uninsured Deposits / TA 0.010*** 0.011*** 0.011*** 0.017***
(0.001) (0.001) (0.001) (0.002)
Loans / TA 0.008*** 0.008*** 0.008*** 0.013***
(0.000) (0.000) (0.000) (0.001)
Equity / TA 0.006*** 0.008*** 0.007*** 0.010***
(0.002) (0.002) (0.002) (0.004)
FHLB / TA 0.004** -0.009** -0.008** -0.015
(0.002) (0.004) (0.004) (0.013)
Log(Total Assets) -0.001*** -0.001*** -0.001*** -0.002***
(0.000) (0.000) (0.000) (0.000)
Other Borrowing / TA 0.013*** 0.012*** 0.025**
(0.004) (0.004) (0.013)
Fed Funds Purchased / TA -0.013
(0.010)
Repo / TA -0.000
(0.004)
factor(size_quartile)2 -0.000
(0.001)
factor(size_quartile)3 -0.001
(0.001)
factor(size_quartile)4 -0.001
(0.001)
Insured Deposits / TA:factor(size_quartile)2 -0.001
(0.001)
Insured Deposits / TA:factor(size_quartile)3 0.000
(0.001)
Insured Deposits / TA:factor(size_quartile)4 -0.000
(0.001)
(Intercept) 0.012***
(0.002)
Num.Obs. 75987 75987 75987 4737
R2 0.648 0.649 0.651 0.357
R2 Adj. 0.648 0.649 0.651 0.356
R2 Within 0.243 0.245 0.249
R2 Within Adj. 0.243 0.245 0.249
Std.Errors by: idrssd by: idrssd by: idrssd Heteroskedasticity-robust
FE: period X X X

5.2 Interpreting the Coefficients

cat("================================================================\n")
## ================================================================
cat("  HEDONIC DEPOSIT COST COEFFICIENTS\n")
##   HEDONIC DEPOSIT COST COEFFICIENTS
cat("================================================================\n\n")
## ================================================================
# Use Specification 2 (full controls + FE) as the main specification
main_reg <- reg2
b_ins  <- coef(main_reg)["insured_dep_to_ta"]
b_unin <- coef(main_reg)["uninsured_dep_to_ta"]
se_ins  <- sqrt(vcov(main_reg)["insured_dep_to_ta","insured_dep_to_ta"])
se_unin <- sqrt(vcov(main_reg)["uninsured_dep_to_ta","uninsured_dep_to_ta"])

cat(sprintf("  β_insured   = %.5f (SE = %.5f)  → %.3f%% cost per dollar of insured deposits\n",
            b_ins, se_ins, b_ins * 100))
##   β_insured   = 0.01394 (SE = 0.00104)  → 1.394% cost per dollar of insured deposits
cat(sprintf("  β_uninsured = %.5f (SE = %.5f)  → %.3f%% cost per dollar of uninsured deposits\n",
            b_unin, se_unin, b_unin * 100))
##   β_uninsured = 0.01082 (SE = 0.00116)  → 1.082% cost per dollar of uninsured deposits
# Average deposit cost
avg_ins_share  <- mean(df_reg$insured_dep_to_ta / (df_reg$insured_dep_to_ta + df_reg$uninsured_dep_to_ta), na.rm = TRUE)
avg_unin_share <- 1 - avg_ins_share

c_avg <- b_ins * avg_ins_share + b_unin * avg_unin_share
cat(sprintf("\n  Average deposit cost c = %.5f (%.3f%%)\n", c_avg, c_avg * 100))
## 
##   Average deposit cost c = 0.01314 (1.314%)
cat(sprintf("  (Weighted by: %.1f%% insured, %.1f%% uninsured)\n",
            avg_ins_share * 100, avg_unin_share * 100))
##   (Weighted by: 74.2% insured, 25.8% uninsured)
cat(sprintf("\n  DSSW reference: ~1.317%% average cost per dollar of deposits\n"))
## 
##   DSSW reference: ~1.317% average cost per dollar of deposits

6 Bank-Level Deposit Costs

# ==============================================================================
# Compute bank-level deposit cost c_i for each bank at 2022Q4
# c_i = β_insured × (InsuredDep_i / TotalDep_i) + β_uninsured × (UninsuredDep_i / TotalDep_i)
#
# This is the per-dollar non-interest cost of maintaining deposits
# ==============================================================================

compute_bank_costs <- function(df_base, reg_model) {
  b_ins  <- coef(reg_model)["insured_dep_to_ta"]
  b_unin <- coef(reg_model)["uninsured_dep_to_ta"]

  df_base %>%
    mutate(
      ins_share_dep  = insured_deposit / (insured_deposit + uninsured_deposit),
      unin_share_dep = uninsured_deposit / (insured_deposit + uninsured_deposit),

      # Per-dollar deposit cost (weighted by bank's deposit composition)
      deposit_cost_insured   = b_ins,
      deposit_cost_uninsured = b_unin,
      deposit_cost_weighted  = b_ins * ins_share_dep + b_unin * unin_share_dep,

      # Total cost of deposits = interest cost + non-interest cost
      interest_cost_per_dep  = ifelse(total_deposit > 0,
        total_interest_on_deposits_calc / total_deposit, NA_real_),
      total_deposit_cost     = interest_cost_per_dep + deposit_cost_weighted
    )
}

# Compute for all periods
df_costs <- compute_bank_costs(df, main_reg)

# Summary by period
cost_summary <- df_costs %>%
  group_by(period) %>%
  summarise(
    N = n(),
    mean_c   = round(mean(deposit_cost_weighted, na.rm = TRUE), 5),
    med_c    = round(median(deposit_cost_weighted, na.rm = TRUE), 5),
    sd_c     = round(sd(deposit_cost_weighted, na.rm = TRUE), 5),
    mean_int = round(mean(interest_cost_per_dep, na.rm = TRUE), 5),
    mean_tot = round(mean(total_deposit_cost, na.rm = TRUE), 5),
    .groups = "drop"
  )

kbl(cost_summary, format = "html",
    col.names = c("Period","N","Mean c","Median c","SD c",
                  "Mean Int. Cost","Mean Total Cost"),
    caption = "Bank-Level Deposit Costs by Period") %>%
  kable_styling(bootstrap_options = c("striped","condensed"), full_width = FALSE)
Bank-Level Deposit Costs by Period
Period N Mean c Median c SD c Mean Int. Cost Mean Total Cost
2021Q1 5027 0.01316 0.01323 0.00045 0.00097 0.01413
2021Q2 4998 0.01313 0.01321 0.00046 0.00182 0.01496
2021Q3 4962 0.01311 0.01319 0.00046 0.00258 0.01569
2021Q4 4887 0.01308 0.01317 0.00047 0.00322 0.01631
2022Q1 4844 0.01308 0.01315 0.00046 0.00065 0.01373
2022Q2 4812 0.01307 0.01315 0.00046 0.00134 0.01441
2022Q3 4765 0.01307 0.01314 0.00046 0.00233 0.01540
2022Q4 4737 0.01307 0.01314 0.00046 0.00394 0.01701
2023Q1 4712 0.01312 0.01318 0.00044 0.00244 0.01556
2023Q2 4681 0.01315 0.01320 0.00043 0.00569 0.01884
2023Q3 4641 0.01315 0.01321 0.00043 0.00955 0.02270
2023Q4 4630 0.01314 0.01320 0.00043 1.73535 1.74849
2024Q1 4610 0.01315 0.01322 0.00043 0.00479 0.01795
2024Q2 4584 0.01315 0.01322 0.00043 0.00985 0.02301
2024Q3 4566 0.01314 0.01321 0.00043 0.01497 0.02812
2024Q4 4533 0.01313 0.01319 0.00043 0.01985 0.03298

6.1 Distribution of Deposit Costs

df_plot <- df_costs %>% filter(period == "2022Q4")

p1 <- ggplot(df_plot, aes(x = deposit_cost_weighted * 100)) +
  geom_histogram(bins = 50, fill = "steelblue", alpha = 0.7) +
  geom_vline(xintercept = mean(df_plot$deposit_cost_weighted, na.rm = TRUE) * 100,
             color = "red", linetype = "dashed", linewidth = 1) +
  labs(title = "Distribution of Deposit Costs (c), 2022Q4",
       x = "Non-Interest Deposit Cost (% of deposits)", y = "Count") +
  theme_minimal()

p2 <- ggplot(df_plot, aes(x = total_deposit_cost * 100)) +
  geom_histogram(bins = 50, fill = "darkgreen", alpha = 0.7) +
  geom_vline(xintercept = mean(df_plot$total_deposit_cost, na.rm = TRUE) * 100,
             color = "red", linetype = "dashed", linewidth = 1) +
  labs(title = "Total Deposit Cost (Interest + Non-Interest), 2022Q4",
       x = "Total Cost (% of deposits)", y = "Count") +
  theme_minimal()

library(patchwork)
p1 + p2
## Warning: Removed 52 rows containing non-finite outside the scale range (`stat_bin()`).
## Removed 52 rows containing non-finite outside the scale range (`stat_bin()`).

7 Save Output

# ==============================================================================
# Save bank-level deposit costs for use in main analysis
# ==============================================================================

output_costs <- df_costs %>%
  select(idrssd, period,
         deposit_cost_insured, deposit_cost_uninsured, deposit_cost_weighted,
         interest_cost_per_dep, total_deposit_cost) %>%
  filter(!is.na(deposit_cost_weighted))

write_csv(output_costs, file.path(DATA_PROC, "dssw_deposit_costs.csv"))
cat(sprintf("Saved: dssw_deposit_costs.csv (%d rows, %d banks, %d periods)\n",
    nrow(output_costs), n_distinct(output_costs$idrssd), n_distinct(output_costs$period)))
## Saved: dssw_deposit_costs.csv (75149 rows, 5012 banks, 16 periods)
# Also save regression summary
reg_summ <- tibble(
  specification = c("Basic+FE", "FullControls+FE", "SizeInteraction", "2022Q4Only"),
  beta_insured = sapply(models, function(m) coef(m)["insured_dep_to_ta"]),
  beta_uninsured = sapply(models, function(m) coef(m)["uninsured_dep_to_ta"]),
  se_insured = sapply(models, function(m) sqrt(vcov(m)["insured_dep_to_ta","insured_dep_to_ta"])),
  se_uninsured = sapply(models, function(m) sqrt(vcov(m)["uninsured_dep_to_ta","uninsured_dep_to_ta"])),
  r_squared = sapply(models, function(m) summary(m)$r2),
  n_obs = sapply(models, function(m) m$nobs)
)

write_csv(reg_summ, file.path(DATA_PROC, "dssw_deposit_cost_regression.csv"))
cat("Saved: dssw_deposit_cost_regression.csv\n")
## Saved: dssw_deposit_cost_regression.csv

7.1 Diagnostic: Deposit Cost vs. Deposit Beta

# Merge with deposit betas to check relationship
betas <- read_csv(file.path(DATA_PROC, "dssw_deposit_betas.csv"), show_col_types = FALSE) %>%
  mutate(idrssd = as.character(idrssd)) %>%
  filter(estimation_date == "2022Q4") %>%
  select(idrssd, beta_uninsured)

df_diag <- output_costs %>%
  filter(period == "2022Q4") %>%
  left_join(betas, by = "idrssd") %>%
  filter(!is.na(beta_uninsured), !is.na(deposit_cost_weighted))

cor_val <- cor(df_diag$deposit_cost_weighted, df_diag$beta_uninsured, use = "complete.obs")

ggplot(df_diag, aes(x = beta_uninsured, y = deposit_cost_weighted * 100)) +
  geom_point(alpha = 0.3, size = 1) +
  geom_smooth(method = "lm", color = "red", se = TRUE) +
  labs(title = sprintf("Deposit Cost (c) vs. Deposit Beta (β^U) — ρ = %.3f", cor_val),
       x = "Deposit Beta (β^U)", y = "Non-Interest Deposit Cost (%)") +
  theme_minimal()
## `geom_smooth()` using formula = 'y ~ x'

cat(sprintf("Correlation between c and β^U: %.3f\n", cor_val))
## Correlation between c and β^U: 0.020
cat(sprintf("Banks in diagnostic: %d\n", nrow(df_diag)))
## Banks in diagnostic: 4635

8 Summary

cat("================================================================\n")
## ================================================================
cat("  DEPOSIT COST ESTIMATION — FINAL SUMMARY\n")
##   DEPOSIT COST ESTIMATION — FINAL SUMMARY
cat("================================================================\n\n")
## ================================================================
cat("Methodology: DSSW hedonic regression (net non-interest expense approach)\n")
## Methodology: DSSW hedonic regression (net non-interest expense approach)
cat(sprintf("Estimation sample: %d bank-quarters, %d unique banks\n",
    nrow(df_reg), n_distinct(df_reg$idrssd)))
## Estimation sample: 75987 bank-quarters, 5073 unique banks
cat(sprintf("Main specification: Full Controls + Period FE, clustered SE at bank level\n\n"))
## Main specification: Full Controls + Period FE, clustered SE at bank level
cat("Key Results:\n")
## Key Results:
cat(sprintf("  β_insured   = %.5f (%.3f%% per dollar)\n", b_ins, b_ins * 100))
##   β_insured   = 0.01394 (1.394% per dollar)
cat(sprintf("  β_uninsured = %.5f (%.3f%% per dollar)\n", b_unin, b_unin * 100))
##   β_uninsured = 0.01082 (1.082% per dollar)
cat(sprintf("  Average c   = %.5f (%.3f%%)\n\n", c_avg, c_avg * 100))
##   Average c   = 0.01314 (1.314%)
cat("Output files:\n")
## Output files:
cat("  1. dssw_deposit_costs.csv    — Bank-level deposit costs by period\n")
##   1. dssw_deposit_costs.csv    — Bank-level deposit costs by period
cat("  2. dssw_deposit_cost_regression.csv — Regression coefficient summary\n\n")
##   2. dssw_deposit_cost_regression.csv — Regression coefficient summary
cat("Usage in main analysis (revised_franchise_bulletproof_1.Rmd):\n")
## Usage in main analysis (revised_franchise_bulletproof_1.Rmd):
cat("  Load: deposit_costs <- read_csv('dssw_deposit_costs.csv')\n")
##   Load: deposit_costs <- read_csv('dssw_deposit_costs.csv')
cat("  Merge: left_join(deposit_costs %>% filter(period=='2022Q4'), by='idrssd')\n")
##   Merge: left_join(deposit_costs %>% filter(period=='2022Q4'), by='idrssd')
cat("  Franchise: f = (1-β^U)(r-c) × [1-(1+δ)^{-T}]/δ × D^U/A\n")
##   Franchise: f = (1-β^U)(r-c) × [1-(1+δ)^{-T}]/δ × D^U/A