Code for post on the Normativeness of Democracy

auxiliary functions.R

# Auxiliary functions for working with various surveys
library(reshape2)
library(plyr)
library(foreign)
library(ggplot2)

construct_network <- function(adjacencyMatrix, threshold=0.5, vertexattributes=NULL) {
  stopifnot(require(network))
  adjacencyMatrix <- ifelse(adjacencyMatrix < threshold, 0, adjacencyMatrix)
  net <- as.network(adjacencyMatrix,type="ajacency",directed=FALSE, ignore.eval= FALSE, names.eval="similarity")
  for(i in names(vertexattributes)) {
    #print(as.vector(vertexattributes[ ,i]))
    set.vertex.attribute(net, i, as.vector(vertexattributes[ ,i]))
  }
  net
}

autoplot.network <- function(net, ..., mode="fruchtermanreingold",label="vertex.names",xlab="",ylab="",alpha=NULL, layout.par=NULL) {
  stopifnot(require(network),require(sna))
  d <- as.matrix.network.adjacency(net) # get sociomatrix
  #plotcord <- data.frame(position_matrix) 
  layout.fun <- try(match.fun(paste("gplot.layout.", mode, 
                                    sep = "")), silent = TRUE)
  if (class(layout.fun) == "try-error") 
    stop("Error in gplot: no layout function for mode ", 
         mode)
  plotcord <- data.frame(layout.fun(d, layout.par=layout.par))
  colnames(plotcord) = c("X1","X2")
  edglist <- as.matrix.network.edgelist(net)
  edges <- data.frame(plotcord[edglist[,1],], plotcord[edglist[,2],])
  
  for(i in list.edge.attributes(net)) {
    edges <- cbind(edges,get.edge.attribute(net$mel, i))
    
  }
  
  
  for(i in list.vertex.attributes(net)) {
    plotcord <- cbind(plotcord,get.vertex.attribute(net, i))
    
  }
  names(plotcord)[3:length(names(plotcord))] <- list.vertex.attributes(net)
  colnames(edges) <-  c("X1","Y1","X2","Y2",list.edge.attributes(net))
  
  map <- aes_string(x="X1",y="X2",...)
  
  edgeMap <- aes_string(x="X1", y="Y1", xend = "X2", yend = "Y2",alpha=alpha)
  
  labelMap <- aes_string(x="X1",y="X2",label=label)
  
  pnet <- ggplot()  + 
    geom_segment(mapping=edgeMap, 
                 data=edges,color="grey") +
    geom_point(mapping=map, data=plotcord) + 
    scale_x_continuous(breaks = NULL) + scale_y_continuous(breaks = NULL) +
    geom_text(mapping=labelMap,data=plotcord,color="black") +labs(x=xlab,y=ylab)+theme_bw()
  
  pnet
}

create_ccp_network <- function(data=ccp,vars=c("democ","assem","equal","express","freerel","assoc"),vertexattributes = c( "country.x", "cowcode", "systyear", "mean", "sd", "median", "pct025", "pct975", "cumMean"),threshold=0.5) {
  stopifnot(require(cluster),require(network))
  
  matrixData <- as.data.frame(data[ , vars])
  matrixData <- colwise(as.numeric)(matrixData)
  adjacencyMatrix <- as.matrix(daisy(matrixData, metric="gower"))
  dimnames(adjacencyMatrix)[[1]] <- paste(data[ , vertexattributes[1]])
  dimnames(adjacencyMatrix)[[2]] <- paste(data[ , vertexattributes[1]])
  adjacencyMatrix <- 1 - adjacencyMatrix
  adjacencyMatrix <- ifelse(adjacencyMatrix < threshold, 0, adjacencyMatrix)
  
  net <- as.network(adjacencyMatrix,type="adjacency",directed=FALSE, ignore.eval= FALSE, names.eval="similarity")
  for(i in vertexattributes) {
    set.vertex.attribute(net, i, as.vector(data[ ,i]))
  }
  #print(net)
  net
}

plot_ccp_network <- function(data=ccp,vars=c("democ","assem","equal","express","freerel","assoc"),title="",threshold=0.5, color="mean",size="cumMean",vertexattributes=c( "cowcode", "country.x", "systyear", "mean", "sd", "median", "pct025", "pct975", "cumMean"), label=expression(paste(vertex.names,systyear))) {
  stopifnot(require(ggplot2))
  net <- create_ccp_network(data=data,vars=vars,threshold=threshold, vertexattributes=vertexattributes)
  autoplot(net, color=color,size=size,alpha="similarity",label=label) + labs(alpha="Degree of similarity",color="Mean UDS score \nin 2006", size="Cumulative UDS score\nas of 2006", title=title)
}

find_communities <- function(vars=c("democ","assem","equal","express","freerel","assoc"),data=ccp,threshold=0) {
  stopifnot(require(igraph), require(cluster))
  adjacencyMatrix <- as.matrix(daisy(as.data.frame(data[ ,vars])))
  adjacencyMatrix <- 1 - adjacencyMatrix
  adjacencyMatrix <- ifelse(adjacencyMatrix < threshold, 0, adjacencyMatrix)
  
  dimnames(adjacencyMatrix)[[1]] <- paste(data$country.x)
  dimnames(adjacencyMatrix)[[2]] <- paste(data$country.y)
  g <- graph.adjacency(adjacencyMatrix,mode="undirected",weighted=TRUE)
  groups <- fastgreedy.community(g)
  groups
}

extract_row <- function(df, n=1) {
  df[sample(nrow(df), n), ]
}

compare_from_subset2 <- function(data=data.4,vars=names(data.4[7:295]), subsetVar = "meanCut", n = 2) {
  stopifnot(require(cluster),require(plyr))
  d <- ddply(data, c(subsetVar), extract_row,n=n)
  #   d <- d[ , c("country.x",vars,subsetVar) ]
  #   print(d)
  d <- d[ , vars ]
  
  distMat <- as.matrix(1-daisy(d, metric="gower"))
  dimnames(distMat)[[1]] <- rep(levels(data[ , subsetVar]),each=n)
  dimnames(distMat)[[2]] <- rep(levels(data[ , subsetVar]),each=n)
  distMat
}

calculate_similarities_3 <- function(data=data.4,vars=liberalDemVars, n = 100, subsetVar="meanCut", FUN="mean") {
  meanV <- replicate(n, compare_from_subset2(data=data,vars=vars, subsetVar=subsetVar))
  apply(meanV, c(1,2), FUN=FUN)
  
}

create_network2 <- function(adjacencyMatrix,threshold=0.5) {
  stopifnot(require(network))
  
  adjacencyMatrix <- 1 - adjacencyMatrix
  adjacencyMatrix <- ifelse(adjacencyMatrix < threshold, 0, adjacencyMatrix)
  
  net <- as.network(adjacencyMatrix,type="adjacency",directed=FALSE, ignore.eval= FALSE, names.eval="similarity")
  #print(net)
  net
}

plot.matrix <- function(d, ..., events.vars, mode="fruchtermanreingold",xlab="",ylab="",alpha=NULL, layout.par=NULL,actors.labels="countries",events.labels="characteristics", eventAttributes) {
  stopifnot(require(network),require(sna),require(ggplot2))
  e <- as.matrix(d[ , events.vars])
  e <- ifelse(is.na(e),0,1)
  net <- as.network(e, bipartite="TRUE")
  mapping <- list(...)
  mapping[sapply(mapping, is.null)] <- "NULL"
  #print(mapping)
  parsed <- lapply(mapping, function(x) parse(text = x)[[1]])
  #print(parsed)
  actorAttributes <- d[ , -events.vars ]
  #print(head(actorAttributes))
  #print(head(eventAttributes))
  
  vertex.attr <- rbind(actorAttributes,eventAttributes)
  
  plotcord <- data.frame(gplot.layout.fruchtermanreingold(net, layout.par=layout.par))
  colnames(plotcord) = c("X1","X2")
    
  plotcord <- cbind(plotcord,vertex.attr)
  edglist <- as.matrix.network.edgelist(net)
  edges <- data.frame(plotcord[edglist[,1],], plotcord[edglist[,2],])
  colnames(edges) <-  c("X1","Y1",paste("events",names(vertex.attr),sep=""),"X2","Y2",paste("actors",names(vertex.attr),sep=""))
  #print(head(edges))
  map <- aes_string(x="X1",y="X2",...)
    
  labelMap <- aes_string(x="X1",y="X2",...)
  edgeMap <- aes_string(x="X1", y="Y1", xend = "X2", yend = "Y2")
  
  pnet <- ggplot()  + 
    geom_segment(mapping=edgeMap, 
                 data=edges,color="lightgrey") +
    geom_point(mapping=map, data=plotcord) + 
    scale_x_continuous(breaks = NULL) + scale_y_continuous(breaks = NULL) +
    geom_text(mapping=labelMap,data=plotcord,color="black") +labs(x=xlab,y=ylab)+theme_bw()
  
  pnet
}

extended codes.csv

geocoded_cowcodes.csv

List-of-variables-ccp.csv

Main graphs.R

# Fig 1

library(foreign)
library(plyr)
library(ggplot2)
ccp <- read.dta("../Data/ccpcncv1_0/ccpcnc/ccpcncv1_0.dta")
ccp <- ccp[, -c(grep("article|comments", names(ccp)))]  #discard variables we are not interested in
uds <- read.csv("../Data/uds_summary.csv")
uds <- ddply(uds, .(country), transform, cumMean = cumsum(mean))
ccp <- merge(ccp, subset(uds, year == 2006), by = c("cowcode"))
geocodes <- read.csv("geocoded_cowcodes.csv")  #For making a map, we need latitude and longitude information for all countries in the dataset; this file contains that info
ccp <- merge(ccp, geocodes)
rm(geocodes)

map <- aes_string(x = "lon", y = "lat", color = "mean", size = "cumMean")
data <- subset(ccp, democ != "1. yes")
labelMap <- aes_string(x = "lon", y = "lat", label = expression(paste(country.x, 
    systyear)))
constitutionsMap <- ggplot() + geom_point(data = data, mapping = map) + borders() + 
    scale_color_gradient2() + scale_size(range = c(3, 9))
constitutionsMap + labs(shape = "Does the constitution\nmention democracy?", 
    color = "Mean UDS score \nin 2006 \n(blue= more democratic)", size = "Cumulative UDS score \nas of 2006\n(bigger = more democratic experience)", 
    title = "Fig. 1: Constitutional documents that do NOT mention 'democracy' or 'democratic' as of 2006") + 
    geom_text(data = data, mapping = labelMap)

ggsave("Fig. 1 Constitutional documents that do not mention democracy.png", 
    width = 16, height = 9)

#Correlations between mention of democracy and measure of democracy

cor(colwise(as.numeric)(ccp[, c("democ", "mean", "cumMean")]))

# note that since 'yes' = 1 and 'no' = 2 a positive value for the
# correlation between democ and the average or cumulative UDS score
# implies a negative correlation between being more democratic and
# mentioning democracy in the constitution

#Fig 2
library(reshape2)
liberalDemVars <- c("assem", "equal", "express", "freerel", "assoc")
data <- ccp[, c("country.x", "mean", "cumMean", "systyear", "evntyear", liberalDemVars)]
data <- melt(data, id.vars = c("country.x", "mean", "cumMean", "systyear", "evntyear"))
data$value <- as.numeric(as.factor(data$value))
data <- subset(data, value > 1)
length(unique(data[, "country.x"]))  # Number of countries that are missing at least one liberal freedom in their constitution

liberal.democracy.differences <- qplot(data = data, x = variable, y = reorder(paste(country.x, 
    systyear), value, FUN = "sum"), fill = mean, geom = "tile")
liberal.democracy.differences + labs(y = "Constitution", x = "Missing liberal democracy principle", 
    fill = "Mean UDS score\nas of 2006") + scale_fill_gradient2() + theme_bw()

ggsave("Fig. 2 Countries missing liberal freedoms in their constitutions.png")

#Fig 3

source("auxiliary functions.R")
rights <- c("democ", names(ccp)[which(names(ccp) == "nat"):which(names(ccp) == 
    "arms")])
rights <- rights[!grepl("_|name|age", rights)]
data <- ccp[, c("country.x", "mean", "cumMean", rights)]
data <- melt(data, id.vars = c("country.x", "mean", "cumMean"))
data <- data[grep("1. yes$", data$value), ]
data$value <- as.numeric(as.factor(data$value))
data <- dcast(data, country.x + mean + cumMean + variable ~ variable)
data <- ddply(data, .(variable), transform, variable = c(rep(NA, length(variable) - 
    1), round(length(variable)/183, 2)))


eventAttributes <- data.frame(country.x = rep(NA, length(names(data)) - 4), 
    mean = rep(0, length(names(data)) - 4), cumMean = rep(0, length(names(data)) - 
        4), variable = names(data)[5:length(names(data))])
plot.matrix(data, events.vars = c(5:length(names(data))), eventAttributes = eventAttributes, 
    color = "mean", label = "variable") + scale_color_gradient2() + labs(color = "Mean UDS score in 2006", 
    size = "Cumulative UDS score by 2006", title = "Rights in world constitutions")

ggsave("Fig. 3 Rights in world constitutions.png", width = 15, height = 8)

#Fig 4

source("auxiliary functions.R")
rights <- c("democ", names(ccp)[which(names(ccp) == "nat"):which(names(ccp) == 
    "arms")])
rights <- rights[!grepl("_|name|age", rights)]
data <- ccp[, c("country.x", "mean", "cumMean", rights)]
data <- melt(data, id.vars = c("country.x", "mean", "cumMean"))
data <- ddply(data, .(variable, value), transform, meanUDS = mean(mean), meanCumUDS = mean(cumMean))
data <- ddply(data, .(variable, value), transform, n = length(mean))
data <- data[grep("1. yes$", data$value), ]
varQuestions <- read.csv("List-of-variables-ccp.csv")
data <- merge(data, varQuestions)

library(stringr)
chars.graph <- qplot(data = data, x = reorder(paste(str_wrap(Question, width = 50), 
    " (", variable, ") ", value, sep = ""), n), y = n/183, color = meanUDS) + 
    coord_flip() + geom_hline(yintercept = c(0.5, 0.8), color = "red") + scale_color_gradient2(midpoint = mean(ccp$mean), 
    mid = "grey", high = "blue", low = "red", breaks = c(min(data$meanUDS), 
        mean(ccp$mean), max(data$meanUDS)), labels = c("Less dem than avg", 
        "Avg level of dem in 2006", "More dem than avg"))
chars.graph + labs(x = "Right or provision", y = "Proportion of constitutions endorsing the right", 
    color = "Average UDS \nscore of countries \nendorsing the right")

ggsave("Fig. 4 Rights endorsed in constitutions.png", height = 25, width = 15)

#Fig 5

source("auxiliary functions.R")
rights <- names(ccp)[which(names(ccp) == "nat"):which(names(ccp) == "arms")]
duties <- names(ccp)[which(names(ccp) == "provwork"):which(names(ccp) == "tradeun")]
rights <- rights[!grepl("_|name|age", rights)]

plot_ccp_network(vars = rights, threshold = 0.85, title = "Similarities in rights provisions (85%  or greater similarity)", 
    vertexattributes = c("mean", "cumMean", "systyear", "country.x"), label = expression(abbreviate(country.x))) + 
    scale_color_gradient2(mid = "grey", low = "red") + scale_size(range = c(3, 
    9))
# plot_ccp_network(vars=duties,threshold=0.9, title='Similarities in
# duties provisions (90% or greater
# similarity)',vertexattributes=c('mean','cumMean','systyear','country.x'),label=expression(abbreviate(country.x)))+scale_color_gradient2(mid='grey',low='red')+
# scale_size(range=c(3,9))

ggsave("Fig. 5 Similarities among constitutions 85.png", width = 15, height = 8)

#Calculations of similarities
source("auxiliary functions.R")
ccp$meanCut <- cut(ccp$mean, 3, labels = c("Dictatorship", "Hybrid", "Democracy"))
codes <- read.csv("../Data/extended codes.csv")

names(codes)[which(names(codes) == "country")] <- "country.codes"
names(codes)[which(names(codes) == "ccode")] <- "cowcode"
allProvisions <- names(ccp)[which(names(ccp) == "model"):which(names(ccp) == 
    "achighed")]

data <- merge(ccp, codes, by = c("cowcode"))
calculate_similarities_3(data = data, n = 300, vars = rights)
##              Dictatorship Dictatorship Hybrid Hybrid Democracy Democracy
## Dictatorship       0.0000       0.6298 0.5964 0.6056    0.6074    0.6033
## Dictatorship       0.6298       0.0000 0.6040 0.6001    0.6006    0.6070
## Hybrid             0.5964       0.6040 0.0000 0.6180    0.5850    0.5871
## Hybrid             0.6056       0.6001 0.6180 0.0000    0.5978    0.5856
## Democracy          0.6074       0.6006 0.5850 0.5978    0.0000    0.6022
## Democracy          0.6033       0.6070 0.5871 0.5856    0.6022    0.0000
calculate_similarities_3(data = data, n = 300, vars = liberalDemVars)
##              Dictatorship Dictatorship Hybrid Hybrid Democracy Democracy
## Dictatorship       0.0000       0.7860 0.8693 0.8733    0.8273    0.8360
## Dictatorship       0.7860       0.0000 0.8627 0.8647    0.8113    0.8180
## Hybrid             0.8693       0.8627 0.0000 0.9693    0.8967    0.8953
## Hybrid             0.8733       0.8647 0.9693 0.0000    0.8993    0.9013
## Democracy          0.8273       0.8113 0.8967 0.8993    0.0000    0.8287
## Democracy          0.8360       0.8180 0.8953 0.9013    0.8287    0.0000
calculate_similarities_3(data = data, n = 100, vars = rights, subsetVar = "un_continent_name")
##          Africa Africa Americas Americas   Asia   Asia Europe Europe
## Africa   0.0000 0.6630   0.6214   0.6425 0.6256 0.6284 0.5961 0.6083
## Africa   0.6630 0.0000   0.6203   0.6072 0.6248 0.6172 0.6041 0.5988
## Americas 0.6214 0.6203   0.0000   0.5919 0.5808 0.5892 0.5816 0.5909
## Americas 0.6425 0.6072   0.5919   0.0000 0.5942 0.5894 0.5822 0.5742
## Asia     0.6256 0.6248   0.5808   0.5942 0.0000 0.6086 0.5703 0.5795
## Asia     0.6284 0.6172   0.5892   0.5894 0.6086 0.0000 0.5664 0.5742
## Europe   0.5961 0.6041   0.5816   0.5822 0.5703 0.5664 0.0000 0.5911
## Europe   0.6083 0.5988   0.5909   0.5742 0.5795 0.5742 0.5911 0.0000
## Oceania  0.6695 0.6692   0.6016   0.6228 0.6523 0.6336 0.5769 0.5839
## Oceania  0.6542 0.6670   0.5962   0.6219 0.6394 0.6223 0.5694 0.5741
##          Oceania Oceania
## Africa    0.6695  0.6542
## Africa    0.6692  0.6670
## Americas  0.6016  0.5962
## Americas  0.6228  0.6219
## Asia      0.6523  0.6394
## Asia      0.6336  0.6223
## Europe    0.5769  0.5694
## Europe    0.5839  0.5741
## Oceania   0.0000  0.7417
## Oceania   0.7417  0.0000

data[, allProvisions] <- colwise(as.factor)(data[, allProvisions])
calculate_similarities_3(data = data, n = 100, vars = allProvisions, subsetVar = "un_continent_name")
##          Africa Africa Americas Americas   Asia   Asia Europe Europe
## Africa   0.0000 0.6787   0.6486   0.6521 0.6641 0.6638 0.6584 0.6563
## Africa   0.6787 0.0000   0.6555   0.6458 0.6679 0.6580 0.6660 0.6613
## Americas 0.6486 0.6555   0.0000   0.6428 0.6410 0.6409 0.6343 0.6254
## Americas 0.6521 0.6458   0.6428   0.0000 0.6401 0.6410 0.6352 0.6333
## Asia     0.6641 0.6679   0.6410   0.6401 0.0000 0.6627 0.6613 0.6513
## Asia     0.6638 0.6580   0.6409   0.6410 0.6627 0.0000 0.6535 0.6496
## Europe   0.6584 0.6660   0.6343   0.6352 0.6613 0.6535 0.0000 0.6734
## Europe   0.6563 0.6613   0.6254   0.6333 0.6513 0.6496 0.6734 0.0000
## Oceania  0.6499 0.6427   0.6473   0.6522 0.6385 0.6322 0.6065 0.6108
## Oceania  0.6308 0.6360   0.6530   0.6382 0.6337 0.6216 0.6106 0.6028
##          Oceania Oceania
## Africa    0.6499  0.6308
## Africa    0.6427  0.6360
## Americas  0.6473  0.6530
## Americas  0.6522  0.6382
## Asia      0.6385  0.6337
## Asia      0.6322  0.6216
## Europe    0.6065  0.6106
## Europe    0.6108  0.6028
## Oceania   0.0000  0.6761
## Oceania   0.6761  0.0000
calculate_similarities_3(data = data, n = 300, vars = allProvisions)

##              Dictatorship Dictatorship Hybrid Hybrid Democracy Democracy
## Dictatorship       0.0000       0.6537 0.6517 0.6520    0.6431    0.6434
## Dictatorship       0.6537       0.0000 0.6505 0.6548    0.6442    0.6416
## Hybrid             0.6517       0.6505 0.0000 0.6638    0.6489    0.6454
## Hybrid             0.6520       0.6548 0.6638 0.0000    0.6537    0.6507
## Democracy          0.6431       0.6442 0.6489 0.6537    0.0000    0.6547
## Democracy          0.6434       0.6416 0.6454 0.6507    0.6547    0.0000