#                   PBC IN LATAM: A DYNAMIC PANEL ANALYSIS
#                        Author: Erick Alvarez Barreno
#                                 June 2021

# Libraries:
library(openxlsx)
library(tidyverse)
library(ggplot2)
library(ggpubr)
library(stargazer)
library(plm)

# Information for the Dynamic Panel Analysis:
## https://cran.r-project.org/web/packages/plm/plm.pdf


# Database available here:
## https://drive.google.com/drive/folders/1-GPe8CjeUvPYNdzYO4UUmWBZGpH4rYYX?usp=sharing


# Import the database
PBC_Database<- read.xlsx("C:/Users/User/Desktop/Revisión Artículo Antioquia/PBC_Database.xlsx")
glimpse(PBC_Database)


# Create the log of Realgdpcapg:
PBC_Database <- PBC_Database %>% 
  mutate(log_realgdpcapg = log(Realgdpcapg)) %>% 
  mutate(Unemp = as.double(Unemp))

# Filter the US out the database:
  PBC_Database <- PBC_Database %>% 
  filter(COUNTRY != "United States")

# Create levels of Presidential approval (high approval as base category):
PBC_Database <- PBC_Database %>% 
  mutate(Low_execapp = ifelse(test = ExecApp < 40, yes = 1, no = 0),
         Low_execapp = ifelse(test = ExecApp >= 40, yes = 0, no = 1),
         Med_execapp = ifelse(test = ExecApp >= 40 & ExecApp <= 60, yes = 1, no = 0),
         Med_execapp = ifelse(test = ExecApp < 40 | ExecApp > 60, yes = 0, no = 1))

# Create one lag and one lead for electoral years to test what government year matters:
PBC_Database <- PBC_Database %>% 
  mutate(LagElect_1 = ifelse(test = dplyr::lead(Elect, 1), yes = 1, no = 0),
         LeadElect_1 = ifelse(test = dplyr::lag(Elect, 1), yes = 1, no = 0))


# Table 1. Descriptive Statistics
stargazer(select(PBC_Database, Elect, LeadElect_1, LagElect_1, CurrExp,
                 CapExp, ExecApp, Taxes, Unemp, Gdpg, log_realgdpcapg),
          type = "text",
          summary.stat = c("n", "mean", "sd", "min", "max"),
          digits = 2,
          header = FALSE, 
          covariate.labels = c("Año electoral",
                               "Año poselectoral",
                               "Año preelectoral",
                               "Gasto corriente (% del PIB)",
                               "Gasto de capital (% del PIB)",
                               "Nivel de aprobación presidencial",
                               "Ingresos tributarios (% del PIB)",
                               "Desempleo (% total fuerza laboral)",
                               "Tasa crecimiento del PIB",
                               "Log PIB per cápita real"),
            style = "qje", out = "Table_1.doc")

# Treat electoral dummies as factors:
PBC_Database <- PBC_Database %>% 
  mutate(Elect = as.factor(Elect),
         LagElect_1 = as.factor(LagElect_1),
         LeadElect_1 = as.factor(LeadElect_1))

# Boxplot: Electoral dynamics and government spending
A <- PBC_Database %>% 
  filter(!is.na(Elect)) %>% 
  ggplot(aes(x = Elect, y = CurrExp, fill = factor(Elect))) +
  geom_boxplot() +
  theme_classic() +
  geom_jitter(width = 0.1, alpha = 0.2) +
  guides(fill = FALSE) +
  scale_fill_manual(values = alpha(c("grey40", "grey40"), 0.20)) +
  labs(y = "",
       x = "Año electoral")

B <- PBC_Database %>% 
  filter(!is.na(LagElect_1)) %>% 
  ggplot(aes(x = LagElect_1, y = CurrExp, fill = factor(LagElect_1))) +
  geom_boxplot() +
  theme_classic() +
  geom_jitter(width = 0.1, alpha = 0.2) +
  guides(fill = FALSE) +
  scale_fill_manual(values = alpha(c("grey40", "grey40"), 0.20)) +
  labs(y = "Gasto corriente (% del PIB)", x = "Año preelectoral")                    

C <- PBC_Database %>% 
  filter(!is.na(LeadElect_1)) %>% 
  ggplot(aes(x = LeadElect_1, y = CurrExp, fill = factor(LeadElect_1))) +
  geom_boxplot() +
  theme_classic() +
  geom_jitter(width = 0.1, alpha = 0.2) +
  guides(fill = FALSE) +
  scale_fill_manual(values = alpha(c("grey40", "grey40"), 0.20)) +
  labs(y = "", x = "Año poselectoral")    


## Graph 1: Electoral dynamics and current expenditure
Graph_1 <- ggarrange(B, A, C + font("x.text", size = 10),
                   ncol = 3, nrow = 1)
                 
Graph_1 


D <- PBC_Database %>% 
  filter(!is.na(Elect)) %>% 
  ggplot(aes(x = Elect, y = CapExp, fill = factor(Elect))) +
  geom_boxplot() +
  theme_classic() +
  geom_jitter(width = 0.1, alpha = 0.2) +
  guides(fill = FALSE) +
  scale_fill_manual(values = alpha(c("grey40", "grey40"), 0.20)) +
  labs(y = "",
       x = "Año electoral")

E <- PBC_Database %>% 
  filter(!is.na(LagElect_1)) %>% 
  ggplot(aes(x = LagElect_1, y = CapExp, fill = factor(LagElect_1))) +
  geom_boxplot() +
  theme_classic() +
  geom_jitter(width = 0.1, alpha = 0.2) +
  guides(fill = FALSE) +
  scale_fill_manual(values = alpha(c("grey40", "grey40"), 0.20)) +
  labs(y = "Gasto de capital (% del PIB)", x = "Año pre-electoral")                    

G <- PBC_Database %>% 
  filter(!is.na(LeadElect_1)) %>% 
  ggplot(aes(x = LeadElect_1, y = CapExp, fill = factor(LeadElect_1))) +
  geom_boxplot() +
  theme_classic() +
  geom_jitter(width = 0.1, alpha = 0.2) +
  guides(fill = FALSE) +
  scale_fill_manual(values = alpha(c("grey40", "grey40"), 0.20)) +
  labs(y = "", x = "Año post-electoral")

## Graph 2: Electoral dynamics and capital expenditure
Graph_2 <- ggarrange(E, D, G + font("x.text", size = 10),
                     ncol = 3, nrow = 1)

Graph_2


# Graph 3. Impact of election year on current expenditure
Graph_3 <- PBC_Database %>%
  group_by(COUNTRY, Elect) %>% 
  dplyr::summarise(CurrExp = mean(CurrExp, na.rm = TRUE)) %>% 
  ungroup() %>% 
  reshape2::dcast(COUNTRY ~ Elect, mean, value.var = "CurrExp", na.rm = TRUE) %>% 
  dplyr::rename(Elect_year = "1", Nonelect_year = "0") %>%
  mutate(Dif_elect = Elect_year - Nonelect_year)

currexp_LATAM <- read.xlsx("C:/Users/User/Desktop/Revisión Artículo Antioquia/Mean Latam.xlsx",
                         sheet = 1)

Graph_3 <- rbind(Graph_3, currexp_LATAM)

ggplot(Graph_3, aes(x = reorder(COUNTRY, Dif_elect), y = Dif_elect)) +
  geom_bar(stat = "identity", fill = "grey40", colour = "black",
           alpha = .30) +
  coord_flip() +
  labs(title = "Variación del gasto corriente en porcentaje del PIB",
       x = "",
       y = "Diferencia del gasto corriente (% del PIB) entre años electorales y no electorales") +
  theme_classic()


## Graph 4. Impact of election on capital expenditure
Graph_4 <- PBC_Database %>%
  group_by(COUNTRY, Elect) %>% 
  dplyr::summarise(CapExp = mean(CapExp, na.rm = TRUE)) %>% 
  ungroup() %>% 
  reshape2::dcast(COUNTRY ~ Elect, mean, value.var = "CapExp", na.rm = TRUE) %>% 
  dplyr::rename(Elect_year = "1", Nonelect_year = "0") %>%
  mutate(Dif_elect = Elect_year - Nonelect_year)

capexp_LATAM <- read.xlsx("C:/Users/User/Desktop/Revisión Artículo Antioquia/Mean Latam.xlsx",
                          sheet = 2)

Graph_4 <- rbind(Graph_4, capexp_LATAM)

ggplot(Graph_4, aes(x = reorder(COUNTRY, Dif_elect), y = Dif_elect)) +
  geom_bar(stat = "identity", fill = "grey40", colour = "black",
           alpha = .30) +
  coord_flip() +
  labs(title = "Variación del gasto de capital en porcentaje del PIB",
       x = "",
       y = "Diferencia en el gasto de capital (% del PIB) entre años electorales y no electorales") +
  theme_classic()


# Brief Multicollinearity Diagnostics (problems with lags of the DV):
library(mctest)
X <- lm(formula = CurrExp ~ Elect + LagElect_1 + LeadElect_1 + 
          CurrExplag1 + CurrExplag2 + Gdpg + log_realgdpcapg,
        data = PBC_Database)  
imcdiag(X)
library(caret)
car::vif(X)


# Panel Data Analysis
pdata <- pdata.frame(PBC_Database, index = c("COUNTRY", "YEAR"))

CurrExp_FE_NApp <- plm(formula = CurrExp ~ Elect + LagElect_1 + LeadElect_1 + 
                         CurrExplag1 + CurrExplag2 + Gdpg + log_realgdpcapg,
                       data = pdata, 
                       index = c("COUNTRY", "YEAR"), effect = "twoways",
                       model = "within")
summary(CurrExp_FE_NApp)

CurrExp_FE_App <- plm(formula = CurrExp ~ Elect + LagElect_1 + LeadElect_1 +
                        Med_execapp + Low_execapp +
                        CurrExplag1 + CurrExplag2 + Gdpg + log_realgdpcapg,
                      data = pdata, index = c("COUNTRY", "YEAR"),
                      model = "within", effect = "twoways")
summary(CurrExp_FE_App)

CapExp_FE_NApp <- plm(formula = CapExp ~ Elect + LagElect_1 + LeadElect_1 + 
                        CapExplag1 + CapExplag2 +
                        Gdpg + log_realgdpcapg, data = pdata, 
                      index = c("COUNTRY", "YEAR"), model = "within",
                      effect = "twoways")
summary(CapExp_FE_NApp)

CapExp_FE_App <- plm(formula = CapExp ~ Elect + LagElect_1 + LeadElect_1 + 
                       Med_execapp + Low_execapp +
                       CapExplag1 + CapExplag2 + Gdpg + log_realgdpcapg,
                      data = pdata, index = c("COUNTRY", "YEAR"),
                      model = "within", effect = "twoways")
summary(CapExp_FE_App)

## Table 2. Estimated coefficients with fixed effects by country and year
stargazer(CurrExp_FE_NApp, CurrExp_FE_App, CapExp_FE_NApp, CapExp_FE_App,
          covariate.labels = c("Año electoral", "Año pre-electoral",
                               "Año post-electoral", "Aprobación media",
                             "Aprobación baja", "Gasto corriente en t-1",
                             "Gasto corriente en t-2", "Gasto de capital en t-1",
                             "Gasto de capital en t-2", "Tasa de crecimiento del PIB",
                             "Log PIB per cápita real"),
          dep.var.labels = c("Gasto corriente", "Gasto de capital"),
          type = "text", out = "Table_2.doc", align = TRUE, no.space = TRUE,
          omit.stat = c("LL","ser", "f"))


## Wooldridge test for serial correlation (Table 3)
pbgtest(CurrExp_FE_NApp)
pbgtest(CurrExp_FE_App)
pbgtest(CapExp_FE_NApp)
pbgtest(CapExp_FE_App)


## Breusch-Pagan test for Homoscedasticity (Table 4)
library(lmtest)
bptest(CurrExp ~ Elect + LagElect_1 + LeadElect_1 + 
         CurrExplag1 + CurrExplag2 + Gdpg + log_realgdpcapg,
       data = pdata, studentize = FALSE)
bptest(CurrExp ~ Elect + LagElect_1 + LeadElect_1 +
         Med_execapp + Low_execapp +
         CurrExplag1 + CurrExplag2 + Gdpg + log_realgdpcapg,
       data = pdata, studentize = FALSE)
bptest(CapExp ~ Elect + LagElect_1 + LeadElect_1 + 
         Med_execapp + Low_execapp +
         CapExplag1 + CapExplag2 + Gdpg + log_realgdpcapg,
       data = pdata, studentize = FALSE)
bptest(CapExp ~ Elect + LagElect_1 + LeadElect_1 + 
         Med_execapp + Low_execapp +
         CapExplag1 + CapExplag2 + Gdpg + log_realgdpcapg,
       data = pdata, studentize = FALSE)


## Pesaran CD test for cross-sectional dependence (Table 5)
pcdtest(CurrExp_FE_NApp, test = c("cd"))
pcdtest(CurrExp_FE_App, test = c("cd"))
pcdtest(CapExp_FE_NApp, test = c("cd"))
pcdtest(CapExp_FE_App, test = c("cd"))


## Controlling for Heteroscedasticity and autocorrelation (Models 1 and 2)
CurrNApp_Robust <- coeftest(CurrExp_FE_NApp,
                            vcovHC(CurrExp_FE_NApp, method = "arellano"))
CurrApp_Robust <- coeftest(CurrExp_FE_App,
                            vcovHC(CurrExp_FE_NApp, method = "arellano"))

## Controlling for Heteroscedasticity (Models 3 and 4)
CapNApp_Robust <- coeftest(CapExp_FE_NApp, vcovHC)
CapApp_Robust <- coeftest(CapExp_FE_App, vcovHC)


## Table 6. Estimates with Robust Standard Errors
stargazer(CurrNApp_Robust, CurrApp_Robust, CapNApp_Robust, CapApp_Robust,
          covariate.labels = c("Año electoral", "Año pre-electoral",
                               "Año post-electoral", "Aprobación media",
                               "Aprobación baja",
                               "Gasto corriente en t-1",
                               "Gasto corriente en t-2",
                               "Gasto de capital en t-1",
                               "Gasto de capital en t-2",
                               "Tasa de crecimiento del PIB",
                               "Log PIB per cápita real"),
          column.labels = c("Gasto corriente", "Gasto de capital"), 
          column.separate = c(2, 2),
          type = "text", out = "Table_6.doc",
          align = TRUE, no.space = TRUE,
          omit.stat = c("LL","ser","f"))


## Generalized Method of Moments (GMM)
GMM_CurrNApp <- pgmm(CurrExp ~ Elect + LagElect_1 + LeadElect_1 +
                       lag(CurrExp, 1:2) + Gdpg + log_realgdpcapg |
                       lag(CurrExp, 2:99) | lag(Elect, 1:2) + lead(Elect, 2),
                    data = PBC_Database,
                    index = c("COUNTRY", "YEAR"),
                    effect = "individual", model = "onestep",
                    subset = sample == 1,
                    transformation = "d")

summary(GMM_CurrNApp)


GMM_CurrApp <- pgmm(CurrExp ~ Elect + LagElect_1 + LeadElect_1 + 
                      Med_execapp + Low_execapp +
                      lag(CurrExp, 1:2) + Gdpg + log_realgdpcapg | 
                      lag(CurrExp, 2:99) | lag(Elect, 1:2) + lead(Elect, 2),
                    data = PBC_Database,
                    index = c("COUNTRY", "YEAR"),
                    effect = "individual", model = "onestep",
                    subset = sample == 1,
                    transformation = "d") 

summary(GMM_CurrApp)


GMM_CapNApp <- pgmm(CapExp ~ Elect + LagElect_1 + LeadElect_1 +
                      lag(CapExp, 1:2)+ Gdpg + log_realgdpcapg |
                      lag(CapExp, 2:99) | lag(Elect, 1:2) + lead(Elect, 2),
                    data = PBC_Database,
                    index = c("COUNTRY", "YEAR"),
                    effect = "individual", model = "onestep",
                    subset = sample == 1,
                    transformation = "d")

summary(GMM_CapNApp)


GMM_CapApp <- pgmm(CapExp ~ Elect + LagElect_1 + LeadElect_1 +
                     Med_execapp + Low_execapp +
                     lag(CapExp, 1:2) + Gdpg + log_realgdpcapg | 
                     lag(CapExp, 2:99) | lag(Elect, 1:2) + lead(Elect, 2),
                   data = PBC_Database,
                   index = c("COUNTRY", "YEAR"),
                   effect = "individual", model = "onestep",
                   subset = sample == 1,
                   transformation = "d")

summary(GMM_CapApp)


# Table 7. Estimates with the Generalized Method of Moments
stargazer(GMM_CurrNApp, GMM_CurrApp, GMM_CapNApp, GMM_CapApp,
          covariate.labels = c(
            "Año electoral",
            "Año pre-electoral",
            "Año post-electoral",
            "Aprobación media",
            "Aprobación baja",
            "Gasto corriente en t-1",
            "Gasto corriente en t-2",
            "Gasto de capital en t-1",
            "Gasto de capital en t-2",
            "Tasa de crecimiento del PIB",
            "Log PIB per cápita real"),
          column.labels = c("Gasto corriente", "Gasto de capital"), 
          column.separate = c(2, 2),
          type = "text", out = "Table_7.doc", align = TRUE, no.space = TRUE,
          omit.stat = c("LL","ser","f", "N"))


# Appendix
# Gov. Expenditure components and electoral years per country:

## Argentina
Elections_Arg <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Argentina",
         YEAR %in% c(1990:2017))

Argentina_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Argentina",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Arg$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Argentina_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Argentina",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Arg$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Argentina_Curr, Argentina_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Bolivia
Elections_Bol <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Bolivia",
         YEAR %in% c(1990:2017))

Bolivia_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Bolivia",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Bol$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Bolivia_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Bolivia",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Bol$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Bolivia_Curr, Bolivia_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Brazil
Elections_Bra <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Brazil",
         YEAR %in% c(1996:2017))

Brazil_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Brazil",
         YEAR %in% c(1996:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Bra$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 7) +
  theme_classic()

Brazil_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Brazil",
         YEAR %in% c(1996:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Bra$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 7) +
  theme_classic()

ggarrange(Brazil_Curr, Brazil_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Chile
Elections_Chile <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Chile",
         YEAR %in% c(1990:2017))

Chile_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Chile",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Chile$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Chile_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Chile",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Chile$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Chile_Curr, Chile_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Colombia
Elections_Col <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Colombia",
         YEAR %in% c(1990:2017))

Colombia_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Colombia",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Col$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Colombia_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Colombia",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Col$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Colombia_Curr, Colombia_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Costa Rica
Elections_CosRica <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Costa Rica",
         YEAR %in% c(1990:2017))

CosRica_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Costa Rica",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_CosRica$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

CosRica_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Costa Rica",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_CosRica$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(CosRica_Curr, CosRica_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Dominican Republic
Elections_DomRep <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Dominican Republic",
         YEAR %in% c(1990:2017))

DomRep_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Dominican Republic",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_DomRep$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

DomRep_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Dominican Republic",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_DomRep$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(DomRep_Curr, DomRep_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Ecuador
Elections_Ecu <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Ecuador",
         YEAR %in% c(1990:2017))

Ecuador_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Ecuador",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Ecu$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Ecuador_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Ecuador",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Ecu$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Ecuador_Curr, Ecuador_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## El Salvador
Elections_Salv <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "El Salvador",
         YEAR %in% c(1990:2017))

Salvador_Curr <- PBC_Database %>% 
  filter(COUNTRY == "El Salvador",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Salv$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Salvador_Cap <- PBC_Database %>% 
  filter(COUNTRY == "El Salvador",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Salv$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Salvador_Curr, Salvador_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Guatemala
Elections_Guat <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Guatemala",
         YEAR %in% c(1990:2017))

Guatemala_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Guatemala",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Guat$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Guatemala_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Guatemala",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Guat$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Guatemala_Curr, Guatemala_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Haiti
Elections_Haiti <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Haiti",
         YEAR %in% c(1990:2017))

Haiti_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Haiti",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Haiti$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Haiti_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Haiti",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Haiti$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Haiti_Curr, Haiti_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Honduras
Elections_Hond <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Honduras",
         YEAR %in% c(1990:2017))

Hond_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Honduras",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Hond$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Hond_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Honduras",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Hond$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Hond_Curr, Hond_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Mexico
Elections_Mex <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Mexico",
         YEAR %in% c(1990:2017))

Mex_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Mexico",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Mex$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Mex_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Mexico",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Mex$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Mex_Curr, Mex_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Nicaragua
Elections_Nic <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Nicaragua",
         YEAR %in% c(1990:2017))

Nic_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Nicaragua",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Nic$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Nic_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Nicaragua",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Nic$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Nic_Curr, Nic_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Panama
Elections_Pan <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Panama",
         YEAR %in% c(1990:2017))

Pan_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Panama",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Pan$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Pan_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Panama",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Pan$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Pan_Curr, Pan_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Paraguay
Elections_Par <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Paraguay",
         YEAR %in% c(1990:2017))

Par_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Paraguay",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Par$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Par_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Paraguay",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Par$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Par_Curr, Par_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Peru
Elections_Peru <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Peru",
         YEAR %in% c(1990:2017))

Peru_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Peru",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Peru$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Peru_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Peru",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Peru$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Peru_Curr, Peru_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Uruguay
Elections_Uru <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Uruguay",
         YEAR %in% c(1990:2017))

Uru_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Uruguay",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Uru$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

Uru_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Uruguay",
         YEAR %in% c(1990:2017)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Uru$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 11) +
  theme_classic()

ggarrange(Uru_Curr, Uru_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)


## Venezuela
Elections_Ven <- PBC_Database %>% 
  select(COUNTRY, YEAR, Elect) %>% 
  filter(Elect == "1",
         COUNTRY == "Venezuela",
         YEAR %in% c(1990:2014))

Ven_Curr <- PBC_Database %>% 
  filter(COUNTRY == "Venezuela",
         YEAR %in% c(1990:2014)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CurrExp)) +
  geom_vline(xintercept = Elections_Ven$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto corriente (% del PIB)") +
  scale_x_continuous(n.breaks = 9) +
  theme_classic()

Ven_Cap <- PBC_Database %>% 
  filter(COUNTRY == "Venezuela",
         YEAR %in% c(1990:2014)) %>% 
  ggplot(aes(x = YEAR)) +
  geom_line(aes(y = CapExp)) +
  geom_vline(xintercept = Elections_Ven$YEAR,
             color = "blue",
             linetype = 6) +
  labs(x = "", y = "Gasto de capital (% del PIB)") +
  scale_x_continuous(n.breaks = 9) +
  theme_classic()

ggarrange(Ven_Curr, Ven_Cap + font("x.text", size = 10),
          ncol = 1, nrow = 2)