In Martin-Summer-1090/syslosseval: Provide data and functions to support the analysis in Systemic Loss Evaluation

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(syslosseval)
library(tidyverse)
library(xtable)

Generate the stressed data:

stress_data_2020 <- make_stress_data(eba_exposures_2020, eba_impairments_2020, 1, 2019)
data_2020 <- make_state_variables(stress_data_2020)

stress_data_2021 <- make_stress_data(eba_exposures_2020, eba_impairments_2020, 2, 2019)
data_2021 <- make_state_variables(stress_data_2021)

stress_data_2022 <- make_stress_data(eba_exposures_2020, eba_impairments_2020, 3, 2019)
data_2022 <- make_state_variables(stress_data_2022)

Generate the base dataset:

initial_state <- make_table_initial_state(data_2020) %>% 
  add_column(State = "initial_2019") %>%
  add_column(Case = "No_Del") %>% 
  rename(Total_Assets = a_0) %>% 
  rename(CET1 = e_0) %>% 
  rename(Leverage = lambda_0)

stress_state_2020 <- make_table_stress_state(data_2020) %>% 
  add_column(State = "adverse_2020")%>% 
  add_column(Case="No_Del") %>% 
  rename(Total_Assets = a_1) %>% 
  rename(CET1 = e_1) %>% 
  rename(Leverage = lambda_1)

stress_state_2021 <- make_table_stress_state(data_2021) %>% 
  add_column(State = "adverse_2021")%>%
  add_column(Case="No_Del") %>% 
  rename(Total_Assets = a_1) %>% 
  rename(CET1 = e_1) %>% 
  rename(Leverage = lambda_1)

stress_state_2022 <- make_table_stress_state(data_2022) %>% 
  add_column(State = "adverse_2022")%>%
  add_column(Case ="No_Del") %>% 
  rename(Total_Assets = a_1) %>% 
  rename(CET1 = e_1) %>% 
  rename(Leverage = lambda_1)

Unlike in 2016 with the 202 dataset we will have banks in the stress scenario the capital of which is completely wiped out in the stress state. We need to isolate these from the data before going into a deeper analysis of the remaining banks.

fundamental_defaulters_2020 <- stress_state_2020 %>% 
  filter(Leverage > 100)
fundamental_defaulters_2021 <- stress_state_2021 %>% 
  filter(Leverage > 100)
fundamental_defaulters_2022 <- stress_state_2022 %>% 
  filter(Leverage > 100)

Compute the share of the total assets of the fundamentally insolvent banks in the total assets of all 121 banks in the system.

Total assets:

a <- initial_state %>% summarize(Total = sum(Total_Assets))
b <- fundamental_defaulters_2020 %>% summarize(Total_2020 = sum(Total_Assets))
c <- fundamental_defaulters_2021 %>% summarize(Total_2021 = sum(Total_Assets))
d <- fundamental_defaulters_2022 %>% summarize(Total_2022 = sum(Total_Assets))

share_tabel_total_assets <- bind_cols(a,b,c,d) %>% 
  mutate(share_2020 = Total_2020/Total) %>% 
  mutate(share_2021 = Total_2021/Total) %>% 
  mutate(share_2022 = Total_2022/Total)

share_tabel_total_assets

Combine all data in a long dataframe excluding the banks whith CET1=0.

leverage_data_no_del <- bind_rows(initial_state, stress_state_2020, stress_state_2021, stress_state_2022) %>%
  mutate(State = factor(State, levels = c("initial_2019", "adverse_2020", "adverse_2021", "adverse_2022")))

exceeders_no_del <- leverage_data_no_del %>% 
  filter(Leverage >= 33)

exceed_no_del <- ggplot(data=exceeders_no_del, mapping = aes(x=State, fill = State))

exceed_plot_2020 <- exceed_no_del + geom_bar()+
  scale_x_discrete(labels=c("initial_2019" = "2019", "adverse_2020" = "2020",
                              "adverse_2021" = "2021", "adverse_2022" = "2022"))+
  scale_fill_discrete(name = "State", labels = c("Initial state year end 2019", 
                                                 "Adverse scenario 2020", "Adverse scenario 2021",
                                                 "Adverse scenario 2022"))+
  labs(x="Year")+
  labs(y="Number of Banks")+
  coord_flip()

#exceed_plot_2020
ggsave(filename = "../paper/Figures/exceed_plot_2020.png", plot = exceed_plot_2020)

aux1 <- initial_state %>% summarize(Total=sum(CET1))
aux2 <- stress_state_2020 %>% summarize(Total_2020 = sum(CET1))
aux3 <- stress_state_2021 %>% summarize(Total_2021 = sum(CET1))
aux4 <- stress_state_2022 %>% summarize(Total_2022 = sum(CET1))

share_table_total_cet1 <- bind_cols(aux1,aux2,aux3,aux4) %>% 
  mutate(cet1_share_2020 = 1-Total_2020/Total) %>% 
  mutate(cet1_share_2021 = 1-Total_2021/Total) %>% 
  mutate(cet1_share_2022 = 1-Total_2022/Total)

share_table_total_cet1

Fire selling

fix_2020 <- fixed_point_computation_function(mat = data_2020, lb = 33, data_idx = sovereign_bond_indices,
            data_adv = average_daily_volume_sovereign, base_year = 2019, constant = 1.5)

fix_2021 <- fixed_point_computation_function(mat = data_2021, lb = 33, data_idx = sovereign_bond_indices,
            data_adv = average_daily_volume_sovereign, base_year = 2019, constant = 1.5)

fix_2022 <- fixed_point_computation_function(mat = data_2022, lb = 33, data_idx = sovereign_bond_indices,
            data_adv = average_daily_volume_sovereign, base_year = 2019, constant = 1.5)

Make a table with the values of the fixed points

c1 <- fix_2020 %>% select(sec_class, delta_lower) %>% 
  rename(delta_2020 = delta_lower)
c2 <- fix_2021 %>% select(delta_lower) %>% 
  rename(delta_2021 = delta_lower)
c3 <- fix_2022 %>% select(delta_lower) %>% 
  rename(delta_2022 = delta_lower)

fixed_points <- bind_cols(c1,c2,c3)

Write a Latex-table

latex_fixed_points <- xtable(fixed_points)
print(latex_fixed_points, file = "../paper/Tables/fixed_points_2020.txt")

Compare to the maximum impact:

max_delta_2020 <- fix_2020 %>% 
  mutate(rel_2020 = delta_upper/delta_max)

max_delta_2021 <- fix_2021 %>% 
  mutate(rel_2021 = delta_upper/delta_max)

max_delta_2022 <- fix_2022 %>% 
  mutate(rel_2022 = delta_upper/delta_max)

Make a table

aux1 <- max_delta_2020 %>% 
  select(sec_class, delta_max, rel_2020)
aux2 <- max_delta_2021 %>% 
  select(rel_2021)
aux3 <- max_delta_2022 %>% 
  select(rel_2022)

rel_table <- bind_cols(aux1, aux2, aux3)

Make a latex table

latex_rel_table <- xtable(rel_table)
print(latex_rel_table, file = "../paper/Tables/rel_table_2020.txt")

Table of the computation results

stress_state_2020_del <- make_table_final_state(data_2020, fix_2020, 33) %>% 
  add_column(State = "adverse_2020")%>% 
  add_column(Case="Del") %>% 
  rename(Total_Assets = a_1_tau) %>% 
  rename(CET1 = e_1_tau) %>% 
  rename(Leverage = lambda_1_tau)

stress_state_2021_del <- make_table_final_state(data_2021, fix_2021, 33) %>% 
  add_column(State = "adverse_2021")%>%
  add_column(Case="Del") %>% 
  rename(Total_Assets = a_1_tau) %>% 
  rename(CET1 = e_1_tau) %>% 
  rename(Leverage = lambda_1_tau)

stress_state_2022_del <- make_table_final_state(data_2022, fix_2022, 33) %>% 
  add_column(State = "adverse_2022")%>%
  add_column(Case ="Del") %>% 
  rename(Total_Assets = a_1_tau) %>% 
  rename(CET1 = e_1_tau) %>% 
  rename(Leverage = lambda_1_tau)

Bind all data into one frame

leverage_data_del <- bind_rows(stress_state_2020_del, stress_state_2021_del, stress_state_2022_del) %>% 
  mutate(State = factor(State, levels = c("adverse_2020", "adverse_2021", "adverse_2022")))

Exceeders:

exceeders_del <- leverage_data_del %>% 
  filter(Leverage >= 33)

set1 <- filter(exceeders_no_del, State == "adverse_2021")
set2 <- filter(exceeders_del, State == "adverse_2021")

Losses in equity:

relative_cet1_del <- leverage_data_del %>% 
  group_by(State) %>% 
  summarize(Tot = sum(CET1))

total_cet1 <- leverage_data_no_del %>% 
  filter(State == "initial_2019") %>% 
  summarize(Total = sum(CET1))

comp_data_no_del <- leverage_data_no_del %>%
  select(Bank_name, CET1, State, Case) 

aux <- filter(comp_data_no_del, State == "initial_2019") %>% 
  mutate(Case = "Del")

comp_data_del <- bind_rows(aux, leverage_data_del) %>%  
  select(Bank_name, CET1, State, Case) 

We need to transform the data for our computations:

comp_data_no_del_trans <- comp_data_no_del %>% 
  pivot_wider(names_from = "State", values_from= "CET1") %>% 
  mutate("Loss 2020" = 1 - adverse_2020/initial_2019) %>% 
  mutate("Loss 2021" = 1 - adverse_2021/initial_2019) %>% 
  mutate("Loss 2022" = 1 - adverse_2022/initial_2019) %>% 
  select(Bank_name, "Loss 2020", "Loss 2021", "Loss 2022", Case) %>% 
  pivot_longer(!c("Bank_name", "Case"), names_to = "Losses", values_to = "Percent")

comp_data_del_trans <- comp_data_del %>% 
  pivot_wider(names_from = "State", values_from= "CET1") %>% 
  mutate("Loss 2020" = 1 - adverse_2020/initial_2019) %>% 
  mutate("Loss 2021" = 1 - adverse_2021/initial_2019) %>% 
  mutate("Loss 2022" = 1 - adverse_2022/initial_2019) %>% 
  select(Bank_name, "Loss 2020", "Loss 2021", "Loss 2022", Case) %>% 
  pivot_longer(!c("Bank_name", "Case"), names_to = "Losses", values_to = "Percent")

compare_data_cet1 <- bind_rows(comp_data_no_del_trans, comp_data_del_trans)

aux_plot <- ggplot(data = compare_data_cet1, mapping = aes(x = Case, y = Percent, fill = Case))

compare_plot <- aux_plot + geom_boxplot(alpha = 0.5) +
  scale_x_discrete(labels=c("Del" = "Del.", "No_Del" = "No Del.")) +
  scale_fill_discrete(name = "Case", labels = c("Deleveraging", "No Deleveraging"))+
  scale_y_continuous(labels = scales::percent_format(accuracy = 1))+
  facet_wrap(~Losses)

compare_plot
ggsave(filename = "../paper/Figures/cet1_losses_2020.png", plot = compare_plot)

Martin-Summer-1090/syslosseval documentation built on Dec. 17, 2021, 3:14 a.m.