02_Data_Analysis.Rmd

---
title: "Data Analysis - The Global Diffusion of Work-Injury Insurance: The Role of Spatial Networks and Nation Building"
output: html_document
---

Nate Breznau, breznau.nate@gmail.com
Felix Lanver, felix.lanver@icloud.com

University of Bremen


```{r setup}
rm(list = ls(all = T))

pacman::p_load('netdiffuseR','stargazer','tidyverse','dplyr','countrycode','sandwich','lmtest','jtools','ragg','ggpubr','rvest','nnet','knitr','Hmisc','survminer','survival')

load("data/.Rdata")

knitr::opts_chunk$set(echo = T)
```

## 1. Final Data Touch Ups



```{r final_cleaning, warning = F, message = F}
# remove missing cases
diff_data <- diff_data[!is.na(diff_data$gdp),]
diff_data_socins <- diff_data_socins[!is.na(diff_data_socins$gdp),]
# Recode non-adopters to 2500
diff_data$toa[is.na(diff_data$toa)] <- 2500
diff_data_socins$toa[is.na(diff_data_socins$toa)] <- 2500

# change 'state existed' to, state founded in last 4 years. We also take the year before formation to account for measurement error in law and state formation timing.

# FIRST LAW
diff_data <- diff_data %>%
  group_by(id) %>%
  mutate(existence = ifelse(existence < 1880, 1880, existence),
         existence = ifelse((per - existence) > -2 & (per - existence) < 5, 1, 0))

# SOCINS
diff_data_socins <- diff_data_socins %>%
  group_by(id) %>%
  mutate(existence = ifelse(existence < 1880, 1880, existence),
         year_nation = existence,
         formation = per - year_nation, # for treatment effect plot
         existence = ifelse((per - existence) > -2 & (per - existence) < 5, 1, 0))

treat_plot <- select(diff_data_socins, toa, id, per, formation, year_nation)
colnames(treat_plot) <- c("socins" ,"id","per","formation","year_nation")
treat_plot <- left_join(treat_plot, diff_data, by = c("id","per"))

# Create relative GDP and ag measures, this puts the country in its global position and follows this changing position over time. Without this change, a lower GDP/higher agricutlrue percent simply predicts year (i.e. lower/higher reflets an earlier year). Also create within country GDP (and trim GDP outliers).

diff_data <- diff_data %>%
  group_by(per) %>%
  mutate(gdp_year_mean = mean(gdp10000, na.rm = T),
         ag_year_mean = mean(pct_ag_01_i, na.rm = T)) %>%
  ungroup()

diff_data <- diff_data %>%
  mutate(gdp_rel = gdp10000 - gdp_year_mean,
         ag_rel = pct_ag_01_i - ag_year_mean) %>%
  group_by(id) %>%
  mutate(gdp_id_mean = mean(gdp10000, na.rm = T),
         gdp_within = gdp10000 - gdp_id_mean)

diff_data_socins <- diff_data_socins %>%
  group_by(per) %>%
  mutate(gdp_year_mean = mean(gdp10000, na.rm = T),
         ag_year_mean = mean(pct_ag_01_i, na.rm = T)) %>%
  ungroup()

diff_data_socins <- diff_data_socins %>%
  mutate(gdp_rel = gdp10000 - gdp_year_mean,
         ag_rel = pct_ag_01_i - ag_year_mean) %>%
  group_by(id) %>%
  mutate(gdp_id_mean = mean(gdp10000, na.rm = T),
         gdp_within = gdp10000 - gdp_id_mean)

# create sequence variable (1= no law, 2 = liability/fund, 3 = social insurance)
treat_plot <- treat_plot %>%
  mutate(law_seq = ifelse(per < toa, 1, 2),
         law_seq = ifelse(law_seq == 2 & per >= socins, 3, law_seq))
```

## 2. Main Models

### First Law

We run three discrete-time logistic hazard models:
1. Without main test variable nation formation and without spatial proximity
2. With main test variable nation formation but without spatial proximity
3. All

We tested relative GDP, within-country GDP variance and regular GDP, none produced any effects at all.

```{r main_models, warning = F, message = F}

model1 <- glm(adopted ~ 
              + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + lag_w_expo_culture 
              #+ w_expo_colony
              + lag_w_expo_trade
              + gdp10000
              + regime
              + 0,
      dat    = diff_data, 
      subset = (per <= toa), # - extremely important !
      family = binomial(link="logit"))

model2 <- glm(adopted ~ 
              + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + existence
              + lag_w_expo_culture 
              #+ w_expo_colony
              + lag_w_expo_trade
              + gdp10000
              + regime
              + 0,
              dat    = diff_data, 
              subset = (per <= toa),
              family = binomial(link="logit"))

model3 <- glm(adopted ~ 
                + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + existence
              + lag_w_expo_culture 
              #+ w_expo_colony
              + lag_w_expo_trade
              + lag_w_expo_proximity
              + gdp10000
              + regime
              + 0,
              dat    = diff_data, 
              subset = (per <= toa),
              family = binomial(link="logit"))

model3a <- glm(adopted ~ 
                + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + existence
              + lag_w_expo_culture 
              #+ w_expo_colony
              + lag_w_expo_trade
              + lag_w_expo_proximity
              + gdp10000
              + gdp10000*existence
              + regime
              + 0,
              dat    = diff_data, 
              subset = (per <= toa),
              family = binomial(link="logit"))

#summary(model2)
#exp(coef(model2))
stargazer(model1,model2,model3,
          type = "text", 
          title = "Discrete Time Hazard/Diffusion Models of /nWork-Injury Law, 1880-2010 in 164 countries",
          dep.var.labels=c("First Work-Injury Law"),
          covariate.labels=c("(1880-1902)", 
                             "(1903-1928)",
                             "(1929-1954)",
                             "(1955-1979)",
                             "(1980-2010)",
                             "State Founded (last 4 years)", 
                             "Network Exposure: Culture (t-1)",
                             #"Network Exposure: Colonial",
                             "Network Exposure: Trade (t-1)",
                             "Network Exposure: Spatial Proximity",
                             "GDPpc 10k USD",
                             "Democratization" 
                             ),
          out="results/Tbl1.htm")

# knitr::include_graphics("results/Tbl1.htm")
```

### Social Ins

```{r main_models_socins, warning = F, message = F}

model1_socins <- glm(adopted ~ 
              + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + lag_w_expo_culture 
              #+ w_expo_colony
              + lag_w_expo_trade
              + gdp10000
              + regime
              + 0,
      dat    = diff_data_socins, 
      subset = (per <= toa), # - extremely important !
      family = binomial(link="logit"))

model2_socins <- glm(adopted ~ 
                + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + existence
              + lag_w_expo_culture 
              #+ w_expo_colony
              + lag_w_expo_trade
              + gdp10000
              + regime
              + 0,
              dat    = diff_data_socins, 
              subset = (per <= toa),
              family = binomial(link="logit"))

model3_socins <- glm(adopted ~ 
                + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + existence
              + lag_w_expo_culture 
              #+ w_expo_colony
              + lag_w_expo_trade
              + lag_w_expo_proximity
              + gdp10000
              + regime
              + 0,
              dat    = diff_data_socins, 
              subset = (per <= toa),
              family = binomial(link="logit"))
#summary(model2)
#exp(coef(model2))
stargazer(model1_socins,model2_socins,model3_socins,
          type = "text", 
          title = "Discrete Time Hazard/Diffusion Models of /nWork-Injury Law, 1880-2010 in 152 countries",
          dep.var.labels=c("First Work-Injury Social Insurance Law"),
          covariate.labels=c("(1880-1902)", 
                             "(1903-1928)",
                             "(1929-1954)",
                             "(1955-1979)",
                             "(1980-2010)",
                             "State Founded (last 4 years)", 
                             "Network Exposure: Culture (t-1)",
                             #"Network Exposure: Colonial",
                             "Network Exposure: Trade (t-1)",
                             "Network Exposrue: Spatial Proximity",
                             "GDPpc 10k USD",
                             "Democratization" 
                             ),
          out="results/Tbl2.htm")

```



### Coef Plots

We run these at 99% confidence but label it at 95, because in the robust clustered standard errors in Table 1, they only make the 95% CI cutoff (they are close). I haven't found an easy way to use robust se's in a coef_plot so this is the next best option as they are very similar

#### First Law

```{r coefplots, echo = T, message = F, warning = F}

plot1 <- plot_coefs(model2,model3, 
                    ci.level = 0.99,
                    coefs = c("A. Network Exposure:\n    Culture (t-1)" = "lag_w_expo_culture",
                                    #"B. Network Exposure:\n    Colonial" = "w_expo_colony",
                                    "B. Network Exposure:\n    Trade (t-1)" = "lag_w_expo_trade",
                                    "C. Network Exposure:\n    Spatial Proximity" = "lag_w_expo_proximity", 
                                    "D. State Founded\n    (last 4 years)" = "existence",
                                    "E. GDP per cap" = "gdp10000", 
                                    "F. Democratization" = "regime"), colors = "Qual1") +
  annotate(geom = "text", x = 2, y = 4.2, label = "OR ~ 31", size = 3) +
  annotate(geom = "text", x = 2, y = 3, label = "OR ~ 2.2", size = 3) +
  annotate(geom = "text", x = 1, y = 1, label = "OR ~ 1.2", size = 3) +
  coord_cartesian(xlim = c(-0.5,5)) + 
  theme(legend.position = "none",
        axis.text.y = element_text(color = "black", hjust = 0),
        axis.title = element_text(color = "black"),
        title = element_text(color = "black")) +
  labs(title = "                          First Law") +
  xlab(label = "Coefficient + 95%CI")


```

#### Social insurance


```{r coefplots_socins, echo = T, warning = F, message = F}

plot2 <- plot_coefs(model2_socins,model3_socins,
                    ci.level = 0.99,
                    coefs = c("A." = "lag_w_expo_culture", 
                              #"B" = "w_expo_colony", 
                              "B." = "lag_w_expo_trade", 
                              "C." = "lag_w_expo_proximity", 
                              "D." = "existence", 
                              "E." = "gdp10000", 
                              "F." = "regime"), 
                    colors = "Qual1") +
  annotate(geom = "text", x = 2.1, y = 4.2, label = "OR ~ 20", size = 3) +
  annotate(geom = "text", x = 2.4, y = 3, label = "OR ~ 3.6", size = 3) +
  annotate(geom = "text", x = 0.9, y = 1, label = "OR ~ 1.1", size = 3) +
  annotate(geom = "text", x = 3.2, y = 5, label = "OR ~ 5.1", size = 3) +
  coord_cartesian(xlim = c(-0.5,5)) + 
  theme(legend.position = "none",
        axis.text.y = element_text(color = "black", hjust = 0),
        axis.title = element_text(color = "black"),
        title = element_text(color = "black")) +
  labs(title = "    First Social Insurance Law") +
  xlab(label = "Coefficient + 95%CI")


agg_png(filename = "results/Fig2.png", height = 600, width = 1200, res = 144)
ggarrange(plot1, plot2, widths = c(1,0.75))

dev.off()

knitr::include_graphics("results/Fig2.png")


```

### Table of adoption

```{r pcts, warning = F, message = F}
diff_data_ex <- diff_data[diff_data$existence == 1, ] %>%
  subset(!is.na(id)) %>%
  group_by(id) %>%
  mutate(e = ifelse(toa == per, 1, 0)) %>%
  summarise(enacted = max(e, na.rm = T))

diff_data_ex_socins <- diff_data_socins[diff_data_socins$existence == 1, ] %>%
  subset(!is.na(id)) %>%
  group_by(id) %>%
  mutate(e = ifelse(toa == per, 1, 0)) %>%
  summarise(enacted = max(e, na.rm = T))

all <- diff_data_socins %>% group_by(id) %>% summarise(enacted_ever = max(adopted, na.rm=T))
  
```

Of the 150 countries, `r sum(diff_data_ex$enacted)` enacted a first law and `r sum(diff_data_ex_socins$enacted)` enacted a first social insurance during the first 5 years of nation state formation. Of the `r sum(all$enacted_ever)` countries that ever enacted work-injury social insurance up to 2010, `r round(100*(sum(diff_data_ex_socins$enacted) /sum(all$enacted_ever)),1)` did so during state formation.

### Treatment plot

Plot sequences centered at the formation of nation states

```{r nation_plot, echo = T, warning = F, message = F}

# remove missing
treat_plot <- subset(treat_plot, formation != "Inf")
treat_plot$law_seq <- as.factor(as.character(treat_plot$law_seq, levels = c("1","2","3"), labels = c("None","Liability","Social Insurance")))

# order by formation
treat_plot <- treat_plot %>%
  mutate(order = ifelse(per == socins, formation, -1000),
         order = max(order, na.rm = T),
         order = ifelse(order == -1000, 1000, order))

treat_plot <- treat_plot[order(treat_plot$order),]

treat_plot <- treat_plot %>%
  ungroup() %>%
  mutate(n =  row_number()) 

treat_plot <- treat_plot %>%
  group_by(id) %>%
  mutate(n_id = max(n, na.rm =T)) %>%
  ungroup()


agg_png(filename = "results/Fig1.png", height = 600, width = 750, res = 144)
ggplot(treat_plot, aes(formation, rev(n_id), fill = law_seq)) +
  geom_tile() +
  scale_fill_manual(breaks = c("1", "2", "3"),
  values = c("grey25","chocolate4","seagreen4"),
  labels = c("None", "Employer Liability", "Social Insurance"),
  ) +
  ylab(label = "Country") +
  xlab(label = "Time in Years\n(centered by year of state formation)") +
  xlim(-25,75) +
  geom_vline(xintercept = -0.5, size = 1) +
  theme_classic() +
  theme(axis.text.y = element_blank(),
        axis.line = element_blank(),
        axis.ticks = element_blank(),
        legend.title = element_blank(),
        panel.background = element_rect(color = "grey", fill = "grey"))
dev.off()

knitr::include_graphics("results/Fig1.png", dpi = 36)

```
### Sesitivity Analysis

Including Agriculture instead of GDP. This still needs work as the period 1990 to 2010 seems to have mostly been interpolated and therefore, not of much use.

```{r sens}
model3_sens <- glm(adopted ~ 
              + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + existence
              #+ lag_w_expo_culture 
              #+ w_expo_colony
              + lag_w_expo_trade
              + lag_w_expo_proximity
              #+ ag_rel
              + regime
              + 0,
              dat    = diff_data, 
              subset = (per <= toa),
              family = binomial(link="logit"))

model3_socins_sens <- glm(adopted ~ 
                + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + existence
              #+ lag_w_expo_culture 
              #+ w_expo_colony
              + lag_w_expo_trade
              + lag_w_expo_proximity
              + ag_rel
              + regime
              + 0,
              dat    = diff_data_socins, 
              subset = (per <= toa),
              family = binomial(link="logit"))

stargazer(model3_sens,model3_socins_sens,
          type = "text", 
          title = "Discrete Time Hazard/Diffusion Models of /nWork-Injury Law, 1880-2010 in 151 countries",
          dep.var.labels=c("First Work-Injury Social Insurance Law"),

          out="results/Tbl3_ag_sens.htm")
```


### Visualizations

Adoption and Cumulative Adoption over time

Note that here it doesn't matter which diffnet object we use, the underlying adoption data are identical


#### Hazard Rate by Start GDP plot

Create groups within the data. High/Low GDP in 1900 & 2000.

`mean(diff_data$gdp10000[diff_data$per == 1900], na.rm = T)`
[1] 0.1851475
`mean(diff_data$gdp10000[diff_data$per == 2000], na.rm = T)`
[1] 1.214281

```{r net_groups, warning = F, message = F}
gdp_groups <- diff_data %>%
  select(gdp10000, lag_w_expo_proximity, id, per, adopted) %>%
  mutate(gdp_1900 = ifelse(gdp10000 <= 0.2 & per == 1900, 1, 0),
         gdp_2000 = ifelse(gdp10000 <= 1.4 & per == 2000, 1, 0),
         proximity = ifelse(lag_w_expo_proximity <= 0.5 & adopted == 1 & lag(adopted) == 0, 1, 0)) %>%
  group_by(id) %>%
  summarise(gdp1900 = max(gdp_1900, na.rm = T),
            gdp2000 = max(gdp_2000, na.rm = T),
            proxim = max(proximity, na.rm = T))

diff_data_gdp <- left_join(diff_data, gdp_groups, by = "id")
diff_data_socins_gdp <- left_join(diff_data_socins, gdp_groups, by = "id")

# remove NA
diff_data_gdp <- diff_data_gdp[!is.na(diff_data_gdp$adopted),]
diff_data_socins_gdp <- diff_data_socins_gdp[!is.na(diff_data_socins_gdp$adopted),]


# change to survival data, remove objects after adoption
 diff_data_gdp <- diff_data_gdp %>%
  group_by(id) %>%
  mutate(cut = adopted + lag(adopted),
         cut = ifelse(is.na(cut), 0, cut)) %>%
  subset(cut == 1) %>%
  ungroup()

 diff_data_socins_gdp <- diff_data_socins_gdp %>%
  group_by(id) %>%
  mutate(cut = adopted + lag(adopted),
         cut = ifelse(is.na(cut), 0, cut)) %>%
  subset(cut == 1) %>%
  ungroup()

# create Cox regression object
fit1 <- survfit(Surv(per, adopted) ~ gdp1900, data = diff_data_gdp)

fit3 <- survfit(Surv(per, adopted) ~ gdp1900, data = diff_data_socins_gdp)

```


```{r surv1}
agg_png(filename = "results/Fig3_A.png", height = 600, width = 600, res = 144)

ggsurvplot(fit1, data = diff_data_gdp, 
           fun = "event",
           conf.int = T,
           palette = "jco",
           xlim = c(1880,2010),
           break.time.by = 20,
           xlab = "Year",
           legend.labs = c("Start GDP High","Start GDP Low"),
           title = "First Law",
           ylab = "")
dev.off()
knitr::include_graphics("results/Fig3_A.png", dpi = 48)
```



```{r surv3}
agg_png(filename = "results/Fig3_B.png", height = 600, width = 600, res = 144)

ggsurvplot(fit3, data = diff_data_socins_gdp, 
           fun = "event",
           conf.int = T,
           palette = "jco",
           xlim = c(1880,2010),
           break.time.by = 20,
           xlab = "Year",
           legend.labs = c("Start GDP High","Start GDP Low"),
           title = "First Social Insurance",
           ylab = "")
dev.off()
knitr::include_graphics("results/Fig3_B.png", dpi = 48)

```
#### Alternative Adoption Plots

##### First Law

```{r net1_first, echo = T}

plot_adopters(diffnet_proximity, 
              include.legend = FALSE, what = c("adopt", "cumadopt"))
```

##### Social Insurance

```{r net1_socins}
plot_adopters(diffnet_proximity_socins, 
              include.legend = FALSE, what = c("adopt"), ylim = c(0,0.02))
```

```{r net2}

plot_hazard(diffnet_proximity_socins, ylim=c(0,0.2))
```


```{r net3}
plot_infectsuscep(diffnet_trade_socins, logscale = F)
```


```{r net4}
# network threshold: required proportion or number of neighbors that leads you to adopt
plot_threshold(diffnet_trade_socins, undirected = FALSE, vertex.size = 1/5)
```


### Sensitivity Analysis

Colonial Legacies have an influence that, after the colony ends is essentially constant when colonies become independent. 

Thus, the influence of a previous colony that has already ended at that time becomes smaller proportionally to the currently existing colony. When the country then becomes completely independent, the influence no longer differs proportionally from each other. It is thus frozen in its strength. 

Colonial legacies have an overall declining strength on the diffusion process. 
Thus, we calculated a non_normalized version of exposure. Influence of current and past colonies is still weighted and exposure calculated proportionally. However, exposure is constantly declining with more time after independence. Furthermore, we create a situation in which exposure can be larger than 1, depending on the weights of the network ties. 

Then we re-run our main models with each version of colonialism.

```{r sens2}
model3_sens2 <- glm(adopted ~ 
                + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + existence
              + lag_w_expo_culture 
              + non_normalized_w_expo_colony
              + lag_w_expo_trade
              + lag_w_expo_proximity
              + gdp10000
              + regime
              + 0,
              dat    = diff_data, 
              subset = (per <= toa),
              family = binomial(link="logit"))

model3_sens3 <- glm(adopted ~ 
                + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + existence
              + lag_w_expo_culture 
              + w_expo_colony
              + lag_w_expo_trade
              + lag_w_expo_proximity
              + gdp10000
              + regime
              + 0,
              dat    = diff_data, 
              subset = (per <= toa),
              family = binomial(link="logit"))


model3_socins_sens2 <- glm(adopted ~ 
                + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + existence
              + lag_w_expo_culture 
              + non_normalized_w_expo_colony
              + lag_w_expo_trade
              + lag_w_expo_proximity
              + gdp10000
              + regime
              + 0,
              dat    = diff_data_socins, 
              subset = (per <= toa),
              family = binomial(link="logit"))

model3_socins_sens3 <- glm(adopted ~ 
                + t0_22 
              + t23_48 
              + t49_74 
              + t75_99 
              + t100_130  
              + existence
              + lag_w_expo_culture 
              + w_expo_colony
              + lag_w_expo_trade
              + lag_w_expo_proximity
              + gdp10000
              + regime
              + 0,
              dat    = diff_data_socins, 
              subset = (per <= toa),
              family = binomial(link="logit"))
#summary(model2)
#exp(coef(model2))
stargazer(model3_sens2, model3_sens3, model3_socins_sens2, model3_socins_sens3,
          type = "text", 
          title = "Discrete Time Hazard/Diffusion Models of /nWork-Injury Law, 1880-2010 in 151 countries",
          out = "results/Tbl_sens_non.htm")

```

## Corrected Standard Errors


```{r se_corrections}
# ---------------------------------------------
# --- Calculate the  corrected standard errors
# --- the cultural spheres network
# ---------------------------------------------

# Standard error correction: Huber-White standard errors are calculated AFTER the model
# and used to re-calculate the significance

# step 1: calculate corrected standard errors and save output as an object: 


m1 <- lmtest::coeftest(model1, vcov = vcovCL(model1, type="HC3", cluster =~ cluster_id))
# unclear error taking place here

m2 <- lmtest::coeftest(model2, vcov = vcovCL(model2, type="HC3", cluster=~ cluster_id))

m3 <- lmtest::coeftest(model3, vcov = vcovCL(model3, type="HC3", cluster=~ cluster_id))

m3x <- lmtest::coeftest(model3_sens2, vcov = vcovCL(model3_sens2, type="HC3", cluster=~ cluster_id))

m3xx <- lmtest::coeftest(model3_sens3, vcov = vcovCL(model3_sens3, type="HC3", cluster=~ cluster_id))

m4 <- lmtest::coeftest(model1_socins, vcov = vcovCL(model1_socins, type="HC3", cluster=~ cluster_id))

m5 <- lmtest::coeftest(model2_socins, vcov = vcovCL(model2_socins, type="HC3", cluster=~ cluster_id))

m6 <- lmtest::coeftest(model3_socins, vcov = vcovCL(model3_socins, type="HC3", cluster=~ cluster_id))

m6x <- lmtest::coeftest(model3_socins_sens2, vcov = vcovCL(model3_socins_sens2, type="HC3", cluster=~ cluster_id))

m6xx <- lmtest::coeftest(model3_socins_sens3, vcov = vcovCL(model3_socins_sens3, type="HC3", cluster=~ cluster_id))
# step 2: run stargazer3 with the following specification: 

# 1. odds.ratios = T 

# 2. stargazer3(list(m1, m2)...) this is the output of the coeftest function with the corrected
# standard erors!! NOT THE GLM OUTPUT !!

# 3. origin_model: GLM output. This is nessesary to add the model fit statistics to the table.


# BEWARE of the covariate label order!! All your models need the independent variables
# in the same order. If you have multiple models, add the ALL independent variable names to the function call below under covariate.labels!!
# If one model has less variables, this space will be empty only for that model.

# cheatsheets for starger modification: https://www.jakeruss.com/cheatsheets/stargazer/
# these work with stargazer 3 as well. 

# change the title, the dep.var.labels and the covariate.labels accordingly
```

### Appendix Table 1

```{r se_corrections1}
stargazer3(list(m1, m2, m3, m3x, m3xx), odds.ratios = T, origin_model = list(model1, model2, model3, model3_sens2, model3_sens3), type = "text", se = list(NA, NA, NA, NA, NA),
           title = "Diffusion of Work-Injury Laws, 151 Countries, 1880-2010",
          covariate.labels=c("(1880-1902)", 
                             "(1903-1928)",
                             "(1929-1954)",
                             "(1955-1979)",
                             "(1980-2010)",
                             "State Founded (last 4 years)", 
                             "Network Exposure: Culture (t-1)",
                             "Network Exposure: Colonial (non-norm)",
                             "Network Exposure: Colonial (norm)",
                             "Network Exposure: Trade (t-1)",
                             "Network Exposrue: Spatial Proximity",
                             "GDPpc 10k USD",
                             "Democratization" 
                             ),
          #add.lines = list(c("Countries", length(unique(diff_data$id)), length(unique(diff_data$id)), length(unique(diff_data$id)), length(unique(diff_data$id)), length(unique(diff_data$id)), length(unique(diff_data$id)))),
           out="results/Tbl_Apdx_1.htm")

# ---------------------------------------------
# --- some functions of netdiffuseR, here for 
# --- the cultural spheres network
# ---------------------------------------------

Tbl1 <- as.data.frame(read_html("results/Tbl_Apdx_1.htm") %>% html_table(fill=TRUE))

Tbl1 <- Tbl1[Tbl1$X6 != "" | Tbl1$X5 != "",]

write.csv(Tbl1, file = "results/Tbl1.csv")

```


### Appendix Table 2

```{r se_corrections2}
stargazer3(list(m4, m5, m6, m6x, m6xx), odds.ratios = T, origin_model = list(model1_socins, model2_socins, model3_socins, model3_socins_sens2, model3_socins_sens3), type = "text", se = list(NA, NA, NA, NA, NA),
           title = "Diffusion of Work-Injury Laws, 151 Countries, 1880-2010",
          covariate.labels=c("(1880-1902)", 
                             "(1903-1928)",
                             "(1929-1954)",
                             "(1955-1979)",
                             "(1980-2010)",
                             "State Founded (last 4 years)", 
                             "Network Exposure: Culture (t-1)",
                             "Network Exposure: Colonial (non-norm)",
                             "Network Exposure: Colonial (norm)",
                             "Network Exposure: Trade (t-1)",
                             "Network Exposrue: Spatial Proximity",
                             "GDPpc 10k USD",
                             "Democratization" 
                             ),
          #add.lines = list(c("Countries", length(unique(diff_data$id)), length(unique(diff_data$id)), length(unique(diff_data$id)), length(unique(diff_data$id)), length(unique(diff_data$id)), length(unique(diff_data$id)))),
           out="results/Tbl_Apdx_2.htm")

# ---------------------------------------------
# --- some functions of netdiffuseR, here for 
# --- the cultural spheres network
# ---------------------------------------------

Tbl2 <- as.data.frame(read_html("results/Tbl_Apdx_2.htm") %>% html_table(fill=TRUE))

Tbl2 <- Tbl2[Tbl2$X6 != "" | Tbl2$X5 != "",]

write.csv(Tbl2, file = "results/Tbl2.csv")

```