Análisis Econométrico del Riesgo Financiero Bancario en Corea del Sur
Raquel Martín González
2026-06-06
1. Introducción
Este documento presenta el análisis econométrico del riesgo financiero bancario en Corea del Sur para el periodo 2018-2025. La variable dependiente es el ratio de préstamos morosos (NPL) de los cuatro principales bancos comerciales coreanos: KEB Hana Bank, Kookmin Bank, Shinhan Bank y Woori Bank. Se estima un modelo de datos de panel con efectos fijos individuales y errores estándar robustos de Arellano.
2. Carga de librerías y datos
library(readxl)
library(dplyr)
library(tidyr)
library(zoo)
library(WDI)
library(plm)
library(lmtest)
library(sandwich)
library(ggplot2)3. Estadísticos descriptivos
summary(datos_final4[, c("npl", "roa", "nim", "bis", "provisions", "lcr",
"pib", "inflacion", "tipo_cambio", "deuda_hogares")])## npl roa nim bis
## Min. :0.1800 Min. :0.3500 Min. :1.330 Min. :14.38
## 1st Qu.:0.2575 1st Qu.:0.5400 1st Qu.:1.407 1st Qu.:16.09
## Median :0.3100 Median :0.5900 Median :1.500 Median :17.34
## Mean :0.3099 Mean :0.5842 Mean :1.520 Mean :17.05
## 3rd Qu.:0.3500 3rd Qu.:0.6300 3rd Qu.:1.590 3rd Qu.:17.89
## Max. :0.5200 Max. :0.7600 Max. :1.860 Max. :18.92
## provisions lcr pib inflacion
## Min. :0.3600 Min. : 86.54 Min. :-0.8371 Min. :0.4987
## 1st Qu.:0.4500 1st Qu.: 92.55 1st Qu.: 1.9278 1st Qu.:2.0081
## Median :0.4900 Median :101.81 Median : 1.9278 Median :2.3217
## Mean :0.5061 Mean : 99.52 Mean : 2.1579 Mean :2.4000
## 3rd Qu.:0.5525 3rd Qu.:104.81 3rd Qu.: 2.9203 3rd Qu.:2.6407
## Max. :0.6900 Max. :112.46 Max. : 4.7478 Max. :5.0895
## tipo_cambio deuda_hogares
## Min. :1144 Min. :140.4
## 1st Qu.:1177 1st Qu.:158.6
## Median :1291 Median :160.3
## Mean :1269 Mean :158.7
## 3rd Qu.:1363 3rd Qu.:160.6
## Max. :1363 Max. :163.34. Evolución del NPL por banco
ggplot(datos_final4, aes(x = as.Date(fecha), y = npl, color = banco)) +
geom_line(linewidth = 1) +
labs(title = "Evolución del NPL por banco (2018-2025)",
x = "Trimestre", y = "NPL (%)", color = "Banco") +
theme_minimal()
5. Variables bancarias: distribución por entidad
datos_largo <- datos_final4 %>%
select(banco, npl, roa, nim, bis, provisions, lcr) %>%
pivot_longer(cols = -banco, names_to = "variable", values_to = "valor") %>%
mutate(variable = factor(variable,
levels = c("npl", "roa", "nim", "bis", "provisions", "lcr"),
labels = c("NPL (%)", "ROA (%)", "NIM (%)", "BIS (%)", "Provisions (%)", "LCR (%)")))
ggplot(datos_largo, aes(x = banco, y = valor, fill = banco)) +
geom_boxplot() +
facet_wrap(~variable, scales = "free_y") +
labs(title = "Distribución de variables bancarias por entidad",
x = "", y = "") +
theme_minimal() +
theme(axis.text.x = element_blank(), legend.position = "bottom",
legend.title = element_blank())
6. Variables macroeconómicas: evolución temporal
datos_macro <- datos_final4 %>%
distinct(fecha, pib, inflacion, tipo_cambio, deuda_hogares) %>%
pivot_longer(cols = -fecha, names_to = "variable", values_to = "valor") %>%
mutate(variable = factor(variable,
levels = c("pib", "inflacion", "tipo_cambio", "deuda_hogares"),
labels = c("PIB (%)", "Inflacion (%)", "Tipo de cambio (KRW/USD)", "Deuda hogares (% PIB)")))
ggplot(datos_macro, aes(x = as.Date(fecha), y = valor)) +
geom_line(color = "steelblue", linewidth = 1) +
facet_wrap(~variable, scales = "free_y") +
labs(title = "Evolución de variables macroeconómicas - Corea del Sur (2018-2025)",
x = "", y = "") +
theme_minimal() +
theme(strip.text = element_text(face = "bold"))
7. Modelo de efectos fijos con errores robustos de Arellano
coeftest(modelo_fe4, vcov = vcovHC(modelo_fe4, type = "HC3", method = "arellano"))##
## t test of coefficients:
##
## Estimate Std. Error t value Pr(>|t|)
## roa -0.04812594 0.02420474 -1.9883 0.050636 .
## nim -0.38693011 0.09248549 -4.1837 8.088e-05 ***
## bis 0.01483691 0.00254270 5.8351 1.473e-07 ***
## pib -0.00702416 0.00381821 -1.8396 0.070001 .
## inflacion 0.00789466 0.00827312 0.9543 0.343192
## tipo_cambio -0.00050697 0.00011314 -4.4811 2.781e-05 ***
## deuda_hogares -0.00689285 0.00129611 -5.3181 1.160e-06 ***
## provisions 0.59196860 0.11169375 5.2999 1.245e-06 ***
## lcr 0.00671025 0.00223331 3.0046 0.003672 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 18. Bondad del ajuste
summary(modelo_fe4)## Oneway (individual) effect Within Model
##
## Call:
## plm(formula = npl ~ roa + nim + bis + pib + inflacion + tipo_cambio +
## deuda_hogares + provisions + lcr, data = datos_panel4, effect = "individual",
## model = "within")
##
## Balanced Panel: n = 4, T = 21, N = 84
##
## Residuals:
## Min. 1st Qu. Median 3rd Qu. Max.
## -0.0539997 -0.0199799 0.0010881 0.0190741 0.0723971
##
## Coefficients:
## Estimate Std. Error t-value Pr(>|t|)
## roa -0.04812594 0.07518143 -0.6401 0.5241481
## nim -0.38693011 0.08465370 -4.5707 2.002e-05 ***
## bis 0.01483691 0.00558752 2.6554 0.0097740 **
## pib -0.00702416 0.00559643 -1.2551 0.2135529
## inflacion 0.00789466 0.00732426 1.0779 0.2847365
## tipo_cambio -0.00050697 0.00012224 -4.1474 9.188e-05 ***
## deuda_hogares -0.00689285 0.00169629 -4.0635 0.0001232 ***
## provisions 0.59196860 0.09580239 6.1791 3.614e-08 ***
## lcr 0.00671025 0.00137843 4.8680 6.590e-06 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Total Sum of Squares: 0.46584
## Residual Sum of Squares: 0.070377
## R-Squared: 0.84892
## Adj. R-Squared: 0.82339
## F-statistic: 44.3293 on 9 and 71 DF, p-value: < 2.22e-169. Gráfico de coeficientes
coefs <- coeftest(modelo_fe4, vcov = vcovHC(modelo_fe4, type = "HC3", method = "arellano"))
df_coefs <- data.frame(
variable = rownames(coefs),
coef = coefs[, 1],
se = coefs[, 2]
) %>%
mutate(
lower = coef - 1.96 * se,
upper = coef + 1.96 * se,
significativo = ifelse(lower > 0 | upper < 0, "Significativo", "No significativo"),
variable = factor(variable, levels = rev(variable))
)
ggplot(df_coefs, aes(x = coef, y = variable, color = significativo)) +
geom_point(size = 3) +
geom_errorbarh(aes(xmin = lower, xmax = upper), height = 0.2, linewidth = 0.8) +
geom_vline(xintercept = 0, linetype = "dashed", color = "black") +
scale_color_manual(values = c("Significativo" = "steelblue", "No significativo" = "gray60")) +
labs(title = "Coeficientes estimados con intervalos de confianza al 95%",
subtitle = "Errores estándar robustos de Arellano",
x = "Coeficiente estimado", y = "", color = "") +
theme_minimal() +
theme(legend.position = "bottom")
10. Tests de diagnóstico
# Heterocedasticidad
bptest(modelo_fe4)##
## studentized Breusch-Pagan test
##
## data: modelo_fe4
## BP = 17.724, df = 9, p-value = 0.03852# Autocorrelacion
pbgtest(modelo_fe4)##
## Breusch-Godfrey/Wooldridge test for serial correlation in panel models
##
## data: npl ~ roa + nim + bis + pib + inflacion + tipo_cambio + deuda_hogares + ...
## chisq = 32.673, df = 21, p-value = 0.04997
## alternative hypothesis: serial correlation in idiosyncratic errors11. Gráfico de residuos
datos_final4$residuos <- as.numeric(residuals(modelo_fe4))
ggplot(datos_final4, aes(x = as.Date(fecha), y = residuos, color = banco)) +
geom_line(linewidth = 0.8) +
geom_hline(yintercept = 0, linetype = "dashed", color = "black") +
facet_wrap(~banco, ncol = 2) +
labs(title = "Residuos del modelo por entidad bancaria",
x = "", y = "Residuo") +
theme_minimal() +
theme(legend.position = "none", strip.text = element_text(face = "bold"))