Exploratory_Data_Analysis.Rmd

---
output: html_document
editor_options: 
  chunk_output_type: console
---

```{r setup, include=FALSE, echo=FALSE}

rm(list = ls()) 
Sys.setlocale("LC_ALL","English")
source("libraries/load_packages.R")
OutputDate <- format( as.Date(Sys.Date()) , "%Y%m%d" )

```

## Data Analysis

We gather data from:

1. WB --> JPMorgan EMBI, Debt and GDP
2. IMF --> Reserves
3. S&P --> Credit Ratings
4. St.Louis FED --> US10y and Risk Avertion


```{r DataLoad, echo=FALSE, message=FALSE, warning=FALSE}

RtoImports <- read_csv2("raw_data/reserves_to_imports.csv")
DBSection3 <- read_csv2("outputs/self_insurance_db.csv")
# str(DBSection3)
DBSection3 <- DBSection3 %>% select(-ifscode, -wbcode2, -country_name_ifs)

```


# Total reserves in month of imports:

```{r}

labels.shape <- c("HIC" = 21,
                 "MIC" = 23)

labels.text <- c("HIC" = "High Income  Countries",
                 "MIC" = "Middle Income Countries")

RtoImports %>% 
  filter(wbcode3 %in% c("HIC", "MIC")) %>% 
  ggplot() +
  geom_line(
      mapping  = aes(x = date, y = value, color = wbcode3),
      size     = 1
      # linetype = 'dotted'
    ) +
  geom_point(
      mapping = aes(x = date, y = value, color = wbcode3, shape = wbcode3),
      size    = 1.4,
      # shape   = 21,
      stroke  = 1.1,
      fill    = 'white'
      # alpha   = 0.5
    ) +
  scale_shape_manual(
    values = labels.shape,
    labels = labels.text
    )+
   scale_fill_brewer(
    labels = labels.text,
    type   = 'qual',
    palette = 'Accent'
    )+
  scale_color_brewer(
    labels = labels.text,
    type = 'qual',
    palette = 'Accent'
    )+
  annotate(
    geom  = 'text',
    x     =  ymd('20120101'),
    y     = 4,
    label = 'Rule of thumb (3x)'
  )+
  geom_hline(
    yintercept = 3,
    linetype   = 'dashed',
    color      = 'grey'
  ) + 
  labs(
    title = 'Total Reserves in month of imports',
    caption = 'Source: The World Bank'
  ) +
  theme_bw()+
  theme(
    plot.title         = element_text(hjust = 0.5), 
    axis.ticks         = element_blank(),
    panel.border       = element_blank(),
    axis.title         = element_blank(),
    legend.position    = 'top',
    legend.title       = element_blank(),
    # legend.text        = element_text(color = )
    panel.grid.major.x = element_blank(),
    panel.grid.minor   = element_blank()
  )

```




## Statistics of the master data base:

```{r statistics, message=FALSE, warning=FALSE, include=FALSE}

# Tip:  stargazer only supports data.frame objects (not tibbles)

DBSection3 %>%
  select(spread, us10y, riskaversion, reserves, public_debt, private_debt, rating, MGDP, R_Y, PuD_Y, PrD_Y, D_Y) %>% as.data.frame() %>%
  stargazer(type = 'text',
            title = "Master Data Base, with NAs",
            out=    "outputs/Data_Analysis_Table1.html")

```



```{r boxplots, message=FALSE, warning=FALSE, include=FALSE}

# Tip:  stargazer only supports data.frame objects (not tibbles)

DBSection3 %>%
  select(spread, rating, R_Y, PuD_Y, PrD_Y, D_Y) %>% 
  gather(variable, value) %>%
  ggplot(aes(y= value))+
    geom_boxplot()+
    facet_wrap(~variable, scales = 'free')

```


Let's take a look to the hoarding of reserves, country by country:

```{r}

# names(DBSection3)

country.list <- list(unique(DBSection3$wbcode3))

data <- DBSection3 %>% 
  select(wbcode3, date, reserves) %>% 
  # filter(date >= ymd('20020101')) %>% 
  group_by(wbcode3) %>% 
  mutate(varnorm = reserves - reserves[1]) %>% 
  ungroup(wbcode3)

# min(data$date)

data %>% 
  filter(date %in% c(ymd('20181201'))) %>% 
  mutate(wbcode3 = fct_reorder(wbcode3, varnorm)) %>% 
  na.omit() %>% 
  ggplot()+
  geom_col(
    mapping = aes(x=wbcode3, y=varnorm, fill=wbcode3)
  ) +
  scale_fill_viridis_d(direction = -1)+
  coord_flip()+
  labs(
    title = 'Cumulative Reserves Hoarded',
    subtitle = 'Since 1998' 
  ) +
  theme(
    legend.position = 'none',
    axis.ticks = element_blank(),
    axis.title = element_blank(),
    axis.line.x = element_blank(),
    plot.subtitle = element_text(hjust = 0.5)
  )


aux <- data %>%
  na.omit() %>%
  filter(!wbcode3 %in% c("CHN")) %>%
  select(wbcode3, varnorm) %>%
  group_by(wbcode3) %>% 
  summarise(max = max(varnorm, na.rm = TRUE)) %>% 
  arrange(desc(max)) %>% 
  mutate(size = ifelse(max > 150, 1.2,
                ifelse(wbcode3 == "ARG", 1.2, 0.5)))

sizes <- aux$size
names(sizes) <- aux$wbcode3

ggplot()+
  geom_line(
    data    = na.omit(data) %>% filter(!wbcode3 %in% c("CHN")),
    mapping = aes(x = as.Date(date), y = varnorm, color=wbcode3, size=wbcode3)
  ) +
  geom_label(
    data    = na.omit(data) %>% filter(!wbcode3 %in% c("CHN"), date == max(date)) %>% 
              filter(varnorm > 150 | wbcode3 == "ARG"),
    mapping = aes(x = as.Date(date), y = varnorm, color=wbcode3, label = wbcode3),
    hjust = -0.05 
  ) +
  scale_x_date(
    limits = c(as.Date('1997-12-01'), as.Date('2020-05-01'))
  ) +
  geom_hline(yintercept = 0)+
  scale_size_manual(values = sizes)+
  scale_color_viridis_d(direction = -1) +
  labs(
    title = 'Hoarding of Reserves (Ex-China)',
    subtitle = 'Based 0 in Dec97 - figures in billion of dollars' 
  ) +
  theme(
    legend.position = 'none',
    axis.ticks = element_blank(),
    axis.title = element_blank(),
    axis.line = element_blank(),
    plot.subtitle = element_text(hjust = 0.5)
  )


```

We find that, no surprisingly, China has unproportionally accumulated international reserves.



## A look to Argentina's public debt

```{r ARGENTINA PUBLIC DEBT, message=FALSE, warning=FALSE, include=FALSE}

DBSection3 %>%
  filter( wbcode3 == "ARG", date >= "2010-01-01" ) %>%
  filter (month(date) == 12) %>%
  select ( year, public_debt, PuD_Y, private_debt, PrD_Y) %>%
  mutate (public_debt = round(public_debt,1)) %>%
  # mutate (private_debt = round(private_debt,1)) %>%
  # mutate (total_debt = round(total_debt,1)) %>%
  mutate( PuD_Y = percent(PuD_Y),
          PrD_Y = percent(PrD_Y)) %>%
  # mutate( PrD_Y = percent(PrD_Y)) %>%
  # mutate( D_Y = percent(D_Y)) %>%
  kable(.,
        col.names = c("Year", "US$ Bn", "% GDP", "US$ Bn", "% GDP"),
        caption = "Argentina Exernal Public Debt with private Creditors") %>%
  kable_styling(bootstrap_options = c("striped", "hover", "responsive"),
                full_width = F,
                position = "left") %>%
  add_header_above(c(' ' = 1, 'Public Debt' = 2, 'Private Debt' = 2))
  



```




## TIME TO FILTER!


We filter the master data base with the following date-criteria:

1) Long Span: 2000 a 2017 (2008 & 2009 excluded due to the international crisis)
2) Pre-Crisis: 2000 a 2007
3) Post-Crisis: 2010 a 2017


Then, we clean the data by the following filters:

1) Credit Ratings > 3: To avoid default-spreads
2) Spreads < 1000: To avoid distress countries
3) Exclude China and Lebanon: due to disproportionate reserve and debt ratios.


Statistics of the cleaned-data:



```{r periods selection, echo=FALSE, message=FALSE, warning=FALSE}


all_period <- DBSection3 %>%
              filter(between(year, 2000, 2018) &  year!=2008 &  year!=2009
                     & rating>3 
                     & spread< 1000 
                     & !wbcode3 %in% c("LBN", "CHN") 
                     & lreserves_1 != "NA" &  lpublicdebt != "NA")

pre_crisis <- DBSection3 %>%
              filter(between(year, 2000, 2007) 
                     & rating>3 
                     & spread< 1000
                     & !wbcode3 %in% c("LBN", "CHN")
                     & lreserves_1 != "NA" &  lpublicdebt != "NA")


post_crisis <- DBSection3 %>%
               filter(year >= 2010 
                      & rating>3 
                      & spread< 1000
                      & !wbcode3 %in% c("LBN", "CHN")
                      & lreserves_1 != "NA" &  lpublicdebt != "NA")





stargazer(as.data.frame(all_period[,c(5:12,22:25)]),
          type = "text",
          iqr = FALSE,
          title="Long Span (unbalanced):2000-Present (2008&2009 excluded)"
          )




```


```{r boxplots with filters, message=FALSE, warning=FALSE, include=FALSE}

# Tip:  stargazer only supports data.frame objects (not tibbles)

all_period %>%
  select(spread, rating, R_Y, PuD_Y, PrD_Y, D_Y) %>% 
  gather(variable, value) %>%
  ggplot(aes(y= value))+
    geom_boxplot()+
    facet_wrap(~variable, scales = 'free')

```

## KEY HISTOGRAMS


```{r Histograms, echo=FALSE, message=FALSE, warning=FALSE}


# Sovereign Spread -------

h1a <- ggplot(DBSection3, aes(spread))+
        geom_histogram()+
        geom_vline(aes(xintercept=mean(spread, na.rm=T)),
                  color="blue", linetype="dashed", size=1)+
        ggtitle("Histogram: Spread\nMaster Data")+
        labs(x = "Sovereign Spread")

h1b <- ggplot(all_period, aes(spread))+
       geom_histogram()+
       geom_vline(aes(xintercept=mean(spread, na.rm=T)),
                  color="blue", linetype="dashed", size=1)+
       ggtitle("Histogram: Spread\nWith Filters")+
       labs(x = "Sovereign Spread")


grid.arrange(h1a, h1b, nrow = 1)


# Reserve Ratios -------

h2a <- ggplot(DBSection3, aes(R_Y))+
       geom_histogram()+
       geom_vline(aes(xintercept=mean(R_Y, na.rm=T)),
                  color="blue", linetype="dashed", size=1)+
       ggtitle("Histogram: Reserve Ratio\nMaster Data")+
       labs(x = "Reserve Ratio")


h2b <- ggplot(all_period, aes(R_Y))+
       geom_histogram()+
       geom_vline(aes(xintercept=mean(R_Y, na.rm=T)),
                  color="blue", linetype="dashed", size=1)+
       ggtitle("Histogram: Reserve Ratio\nWith Filters")+
       labs(x = "Reserve Ratio")

grid.arrange(h2a, h2b, nrow = 1)

# Public Debt Spread -------

h3a <- ggplot(DBSection3, aes(PuD_Y))+
       geom_histogram()+
       geom_vline(aes(xintercept=mean(PuD_Y, na.rm=T)),
                  color="blue", linetype="dashed", size=1)+
       ggtitle("Histogram: Public Debt Ratio\nMaster Data")+
       labs(x = "Debt Ratio")


h3b <- ggplot(all_period, aes(PuD_Y))+
       geom_histogram()+
       geom_vline(aes(xintercept=mean(PuD_Y, na.rm=T)),
                  color="blue", linetype="dashed", size=1)+
       ggtitle("Histogram: Public Debt Ratio\nWith Filters")+
       labs(x = "Debt Ratio")
  
grid.arrange(h3a, h3b, nrow = 1)

```







## TIME SPANS AND COUNTIRES


```{r Countries and Period Covered, echo=FALSE, warning=FALSE}



country.list <- as.list(unique(all_period$country_name_wb))
resume2 <- data.frame()

for (j in country.list) {
  
  # j <- "ARG"
  
  df  <- subset(all_period, country_name_wb == j)
  
  obs <- nrow(df)
  begins <- min(df$date)
  ends <- max(df$date)
  
  resume2.j <- data.frame(j, obs, begins, ends)
  
  resume2 <- rbind(resume2, resume2.j)
  
}

colnames(resume2)[1] <- "Country"



kable(resume2,
      col.names = c("Country", "obs", "Begins", "Ends"),
      caption="Countries and Period Covered") %>%
  kable_styling(bootstrap_options = c("striped", "hover", "responsive"),
                full_width = F,
                position = "left")
  




```





Detail public debt with private creditors:


```{r public debt ratios, echo=FALSE, message=FALSE, warning=FALSE}

debt_ratio <- DBSection3 %>%
  select(wbcode3, date, D_Y) %>%
  filter( month(date) == 12 & date > "2013-01-01" ) %>%
  mutate(D_Y  = round(D_Y *100,1), 
         date = format.Date(date, "%b-%y"))

debt_ratio1 <- spread(debt_ratio, date, D_Y)


kable(debt_ratio1) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "responsive"),
                position = "left")


```



## HOW DOES RATIOS OF DEBT AND RESERVES LOOK LIKE?



```{r Scatter of ratios, echo=FALSE, message=FALSE, warning=FALSE}




ratios <- all_period %>%
          select(wbcode3, date, R_Y , PuD_Y ) %>%
          filter( month(date) == 12) %>%
          mutate(D_Y = PuD_Y *100 ,
                 R_Y = R_Y *100 , 
                 date = as.integer(format(date, "%Y")))%>%
          na.omit()


# str(ratios)
max_x <- max(ratios$D_Y,  na.rm = TRUE)
max_y <- max(ratios$R_Y, na.rm = TRUE)
max <- round(max(max_x, max_y),0)+5

# str(ratios)

# ggplot(ratios, aes(D_Y, R_Y, size=0.5, color=wbcode3))+
#   geom_point(alpha = 0.7)+
#   geom_text(aes(label=ifelse(wbcode3=="ARG","ARG",'')), hjust=0, vjust=0)+
#   geom_abline(aes(slope=1, intercept=0))+
#   scale_x_continuous(name="Debt Ratio",limits=c(0,max)) +
#   scale_y_continuous(name="Reserve Ratio",limits=c(0, max))+
#   ggtitle("Ratios de Reservas y Deuda a diciembre de cada año",
#           subtitle = 'Year: {frame_time}')+
#   transition_time(date)+
#   ease_aes('linear')+
#   shadow_wake(wake_length = 0.1, alpha = FALSE)


# ratios %>% 
  # mutate(wbcode3 = fct_lump(wbcode3, ))
ggplot()+
  geom_point(
    data    = ratios, 
    mapping = aes(D_Y/100, R_Y/100, color=wbcode3),
    alpha   = 0.7
    )+
  geom_abline(aes(slope=1, intercept=0))+
  scale_x_continuous(
    name   = "Debt Ratio",
    limits = c(0,max/100),
    labels = percent_format()
    ) +
  scale_y_continuous(
    name   = "Reserve Ratio",
    limits = c(0, max/100),
    labels = percent_format()
    )+
  ggtitle("Debt & Reserves Ratios to GPD")+
  theme_minimal()+
  theme(legend.title = element_blank(),
        legend.justification = 'center')



```




## DEBT IS HIGHER THAN RESERVES, RIGHT?


```{r Excess Debt, echo=FALSE,  warning=FALSE, results="asis"}


country.list <- as.list(unique(all_period$country_name_wb))
resume <- data.frame()

for (j in country.list) {
  
  # j <- "Argentina"
  
  df  <- subset(all_period, country_name_wb == j)
  
  mean.reserves <- mean(df$R_Y, na.rm=TRUE)
  mean.debt <- mean(df$PuD_Y, na.rm=TRUE)
  
  resume.j <- data.frame(j, mean.reserves, mean.debt)
  
  resume <- rbind(resume, resume.j)
  
}

colnames(resume)[1] <- "Country"

resume <- resume %>%
  mutate(excess.debt.bps = round((mean.debt - mean.reserves)*10000)) %>%
  arrange(desc(excess.debt.bps))
  

resume$mean.reserves <- percent(resume$mean.reserves)
resume$mean.debt <- percent(resume$mean.debt)
# str(resume)

for (i in 1:dim(resume)[1]) {
  
  if (is.na(resume$excess.debt.bps[i]))
  {resume$Condition[i] <- NA}
  else
  if (resume$excess.debt.bps[i] > 0) 
  {resume$Condition[i] <- "As Expected"}
  else {resume$Condition[i] <- "Not as Expected"}
} 

kable(resume,
      col.names = c("Country", "Reserve Ratio", "Debt Ratio", "Excess Debt (in bps)", "Condition"),
      caption = "Average Ratios and Excess Debt (2000-2017)") %>%
  kable_styling(bootstrap_options = c("striped", "hover", "responsive"),
                full_width = F,
                position = "left")

# 
# %>%
#   scroll_box(width = "800px", height = "250px")
#   



```



```{r spreads}



DBSection3 %>%
  filter(date >= ymd('20180101'), !wbcode3 == 'VEN') %>%
  group_by(date, rating) %>%
  summarise(spread = mean(spread, na.rm = TRUE)) %>%
  ungroup(date, rating) %>%
  mutate(rating = as.factor(rating)) %>%
  ggplot(aes(x=date))+
  geom_line(aes(y=spread, colour= rating)) +
  geom_line(data = DBSection3 %>% filter(wbcode3 == 'ARG', date >= ymd('20180101')), aes(y=spread), size = 1) +
  geom_line(data = DBSection3 %>% filter(date >= ymd('20180101')), aes(y=(riskaversion*100)-(us10y*100)), size = 1)

```