United Nations Voting Trends Analysis
Kevin Mekulu
November 2, 2017
ABSTRACT:
There are lots of factors that influenced voting trends at the United Nations Council let’s analyze those voting trends and build simple regression models to analyze the overall voting trends which will allow to answer some hidden questions.
1.1 Import Data Analysis
we start by importing the data set and output the data to get an overall idea of the type of data we are dealing with. Then we filter important features in the data such as the name of the countries and their voting percentages(Yes or No)
votes <- readRDS("C:/Users/jkevi_000/Downloads/votes.rds")
# Print the votes dataset
print(votes)## # A tibble: 508,929 x 4
## rcid session vote ccode
## <dbl> <dbl> <dbl> <int>
## 1 46 2 1 2
## 2 46 2 1 20
## 3 46 2 9 31
## 4 46 2 1 40
## 5 46 2 1 41
## 6 46 2 1 42
## 7 46 2 9 51
## 8 46 2 9 52
## 9 46 2 9 53
## 10 46 2 9 54
## # ... with 508,919 more rows# Filter for votes that are "yes", "abstain", or "no"
votes %>% filter(vote <= 3)## # A tibble: 353,547 x 4
## rcid session vote ccode
## <dbl> <dbl> <dbl> <int>
## 1 46 2 1 2
## 2 46 2 1 20
## 3 46 2 1 40
## 4 46 2 1 41
## 5 46 2 1 42
## 6 46 2 1 70
## 7 46 2 1 90
## 8 46 2 1 91
## 9 46 2 1 92
## 10 46 2 1 93
## # ... with 353,537 more rows# Add another %>% step to add a year column
votes %>%
filter(vote <= 3) %>%
mutate(year = 1945 + session)## # A tibble: 353,547 x 5
## rcid session vote ccode year
## <dbl> <dbl> <dbl> <int> <dbl>
## 1 46 2 1 2 1947
## 2 46 2 1 20 1947
## 3 46 2 1 40 1947
## 4 46 2 1 41 1947
## 5 46 2 1 42 1947
## 6 46 2 1 70 1947
## 7 46 2 1 90 1947
## 8 46 2 1 91 1947
## 9 46 2 1 92 1947
## 10 46 2 1 93 1947
## # ... with 353,537 more rows# Load the countrycode package
library(countrycode)
# Convert country code 100
countrycode(100, "cown", "country.name")## [1] "Colombia"# Add a country column within the mutate: votes_processed
votes_processed <- votes %>%
filter(vote <= 3) %>%
mutate(year = session + 1945,
country = countrycode(ccode, "cown", "country.name"))
# Print votes_processed
head(votes_processed)## # A tibble: 6 x 6
## rcid session vote ccode year country
## <dbl> <dbl> <dbl> <int> <dbl> <chr>
## 1 46 2 1 2 1947 United States of America
## 2 46 2 1 20 1947 Canada
## 3 46 2 1 40 1947 Cuba
## 4 46 2 1 41 1947 Haiti
## 5 46 2 1 42 1947 Dominican Republic
## 6 46 2 1 70 1947 Mexico# Find total and fraction of "yes" votes
votes_processed %>% summarize(total = n(),
percent_yes = mean(vote == 1))## # A tibble: 1 x 2
## total percent_yes
## <int> <dbl>
## 1 353547 0.7999248# Change this code to summarize by year
votes_processed %>%
group_by(year) %>%
summarize(total = n(),
percent_yes = mean(vote == 1)) ## # A tibble: 34 x 3
## year total percent_yes
## <dbl> <int> <dbl>
## 1 1947 2039 0.5693968
## 2 1949 3469 0.4375901
## 3 1951 1434 0.5850767
## 4 1953 1537 0.6317502
## 5 1955 2169 0.6947902
## 6 1957 2708 0.6085672
## 7 1959 4326 0.5880721
## 8 1961 7482 0.5729751
## 9 1963 3308 0.7294438
## 10 1965 4382 0.7078959
## # ... with 24 more rows# Summarize by country: by_country
by_country <- votes_processed %>%
group_by(country) %>%
summarize(total = n(),
percent_yes = mean(vote == 1))
# You have the votes summarized by country
by_country <- votes_processed %>%
group_by(country) %>%
summarize(total = n(),
percent_yes = mean(vote == 1))
# Print the by_country dataset
head(by_country)## # A tibble: 6 x 3
## country total percent_yes
## <chr> <int> <dbl>
## 1 Afghanistan 2373 0.8592499
## 2 Albania 1695 0.7174041
## 3 Algeria 2213 0.8992318
## 4 Andorra 719 0.6383866
## 5 Angola 1431 0.9238295
## 6 Antigua and Barbuda 1302 0.9124424# Sort in ascending order of percent_yes
by_country %>% arrange(percent_yes)## # A tibble: 200 x 3
## country total percent_yes
## <chr> <int> <dbl>
## 1 Zanzibar 2 0.0000000
## 2 United States of America 2568 0.2694704
## 3 Palau 369 0.3387534
## 4 Israel 2380 0.3407563
## 5 Federal Republic of Germany 1075 0.3972093
## 6 United Kingdom of Great Britain and Northern Ireland 2558 0.4167318
## 7 France 2527 0.4265928
## 8 Micronesia (Federated States of) 724 0.4419890
## 9 Marshall Islands 757 0.4914135
## 10 Belgium 2568 0.4922118
## # ... with 190 more rows# Now sort in descending order
by_country %>% arrange(desc(percent_yes))## # A tibble: 200 x 3
## country total percent_yes
## <chr> <int> <dbl>
## 1 Sao Tome and Principe 1091 0.9761687
## 2 Seychelles 881 0.9750284
## 3 Djibouti 1598 0.9612015
## 4 Guinea Bissau 1538 0.9603381
## 5 Timor-Leste 326 0.9570552
## 6 Mauritius 1831 0.9497542
## 7 Zimbabwe 1361 0.9493020
## 8 Comoros 1133 0.9470432
## 9 United Arab Emirates 1934 0.9467425
## 10 Mozambique 1701 0.9465021
## # ... with 190 more rows# Filter out countries with fewer than 100 votes
by_country %>%
arrange(percent_yes) %>%
filter(total >= 100)## # A tibble: 197 x 3
## country total percent_yes
## <chr> <int> <dbl>
## 1 United States of America 2568 0.2694704
## 2 Palau 369 0.3387534
## 3 Israel 2380 0.3407563
## 4 Federal Republic of Germany 1075 0.3972093
## 5 United Kingdom of Great Britain and Northern Ireland 2558 0.4167318
## 6 France 2527 0.4265928
## 7 Micronesia (Federated States of) 724 0.4419890
## 8 Marshall Islands 757 0.4914135
## 9 Belgium 2568 0.4922118
## 10 Canada 2576 0.5081522
## # ... with 187 more rows# Define by_year
by_year <- votes_processed %>%
group_by(year) %>%
summarize(total = n(),
percent_yes = mean(vote == 1))1.2 Data Visualization
This will be the most important step of our data exploration.Let’s take a look at a few plots of the voting trends of different countries, especially the percentage of “Yes” votes amongst different countries over time and see what hypothesis we can infer from those graphs.
# Load the ggplot2 package
library(ggplot2)
# Create line plot
ggplot(by_year, aes(x = year, y = percent_yes))+
geom_line()
# Change to scatter plot and add smoothing curve
ggplot(by_year, aes(year, percent_yes)) +
geom_point() + geom_smooth()## `geom_smooth()` using method = 'loess'
Now lets turn our attention to voting trends observed in specific countries rather than the overall voting trend. We will start by looking at the trend for the United Kingdom then pick a few other countries as well.
# Group by year and country: by_year_country
by_year_country <- votes_processed %>%
group_by(year, country) %>%
summarize(total = n(),
percent_yes = mean(vote == 1))
# Print by_year_country
head(by_year_country)## # A tibble: 6 x 4
## # Groups: year [1]
## year country total percent_yes
## <dbl> <chr> <int> <dbl>
## 1 1947 Afghanistan 34 0.3823529
## 2 1947 Argentina 38 0.5789474
## 3 1947 Australia 38 0.5526316
## 4 1947 Belarus 38 0.5000000
## 5 1947 Belgium 38 0.6052632
## 6 1947 Bolivia (Plurinational State of) 37 0.5945946# Create a filtered version: FR_by_year
FR_by_year <- by_year_country %>%
filter(country == "France")
# Line plot of percent_yes over time for France only
ggplot(FR_by_year, aes(year, percent_yes)) +
geom_line()
# Vector of four countries to examine
countries <- c("United States of America", "Cameroon",
"France", "India")
# Filter by_year_country: filtered_4_countries
filtered_4_countries <- by_year_country %>%
filter(country %in% countries)
# Line plot of % yes in four countries
ggplot(filtered_4_countries, aes(x = year, y = percent_yes, color = country)) +
geom_line()
# Vector of six countries to examine
countries <- c("United States", "Cameroon",
"France", "Japan", "Brazil", "India")
# Filtered by_year_country: filtered_6_countries
filtered_6_countries <- by_year_country %>%
filter(country %in% countries)
# Vector of six countries to examine
countries <- c("United States", "Cameroon",
"France", "Japan", "Brazil", "India")
# Filtered by_year_country: filtered_6_countries
filtered_6_countries <- by_year_country %>%
filter(country %in% countries)
# Line plot of % yes over time faceted by country
ggplot(filtered_6_countries, aes(year, percent_yes)) +
geom_line() +
facet_wrap(~ country, scales = "free_y")
1.3 Predicitve Modelling Using Linear Regression MOdels
We will use a linear regression model in order to examine how one variable changes with respect to another by fitting a best fit line. In this specific case we will use our regression line to describe the association between the percentage of “Yes” votes by different countries over time.
# Load the broom package
library(broom)
# Linear regression of percent_yes by year for US
US_by_year <- by_year_country %>%
filter(country == "United States of America")
US_fit <- lm(percent_yes ~ year, US_by_year)
summary(US_fit)##
## Call:
## lm(formula = percent_yes ~ year, data = US_by_year)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.222491 -0.080635 -0.008661 0.081948 0.194307
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 12.6641455 1.8379743 6.890 8.48e-08 ***
## year -0.0062393 0.0009282 -6.722 1.37e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.1062 on 32 degrees of freedom
## Multiple R-squared: 0.5854, Adjusted R-squared: 0.5724
## F-statistic: 45.18 on 1 and 32 DF, p-value: 1.367e-07The estimated slope of our model is -0.006 meaning that the percentage of yes votes by the United States has decreased by a factor of 0.006 over the years. Now we could dig a little deeper to find out why but we don’t have enough data to answer that question. But to verify wether or not the trend is due to chance let’s look at the p-value which is 1.37*10^(-07) <<< 0.05 so clearly our model is significant.
# Create US_tidied
US_tidied <- tidy(US_fit)1.4 Deploying Linear Regression Models for Multiple Countries
Applying linear regression to individual countries would take us forever instead, let’s deploy our model for multiple countries in one go.
# Load the tidyr package
library(tidyr)
# Nest all columns besides country
nested<- by_year_country %>% group_by(country) %>% nest()
# Unnest the data column to return it to its original form
unnest(nested)## # A tibble: 4,744 x 4
## country year total percent_yes
## <chr> <dbl> <int> <dbl>
## 1 Afghanistan 1947 34 0.3823529
## 2 Afghanistan 1949 51 0.6078431
## 3 Afghanistan 1951 25 0.7600000
## 4 Afghanistan 1953 26 0.7692308
## 5 Afghanistan 1955 37 0.7297297
## 6 Afghanistan 1957 34 0.5294118
## 7 Afghanistan 1959 54 0.6111111
## 8 Afghanistan 1961 76 0.6052632
## 9 Afghanistan 1963 32 0.7812500
## 10 Afghanistan 1965 40 0.8500000
## # ... with 4,734 more rows#Perform Linear Regression on each nested data frame
# Load tidyr and purrr
library(tidyr)
library(purrr)
# Perform a linear regression on each item in the data column
nested %>% mutate(model = map(data, ~lm(percent_yes ~ year, data = .)))## # A tibble: 200 x 3
## country data model
## <chr> <list> <list>
## 1 Afghanistan <tibble [34 x 3]> <S3: lm>
## 2 Argentina <tibble [34 x 3]> <S3: lm>
## 3 Australia <tibble [34 x 3]> <S3: lm>
## 4 Belarus <tibble [34 x 3]> <S3: lm>
## 5 Belgium <tibble [34 x 3]> <S3: lm>
## 6 Bolivia (Plurinational State of) <tibble [34 x 3]> <S3: lm>
## 7 Brazil <tibble [34 x 3]> <S3: lm>
## 8 Canada <tibble [34 x 3]> <S3: lm>
## 9 Chile <tibble [34 x 3]> <S3: lm>
## 10 Colombia <tibble [34 x 3]> <S3: lm>
## # ... with 190 more rows# Load the broom package
library(broom)
# Add another mutate that applies tidy() to each model
nested %>% mutate(model = map(data, ~ lm(percent_yes ~ year, data = .)))%>%
mutate(tidied = map(model,tidy))## # A tibble: 200 x 4
## country data model
## <chr> <list> <list>
## 1 Afghanistan <tibble [34 x 3]> <S3: lm>
## 2 Argentina <tibble [34 x 3]> <S3: lm>
## 3 Australia <tibble [34 x 3]> <S3: lm>
## 4 Belarus <tibble [34 x 3]> <S3: lm>
## 5 Belgium <tibble [34 x 3]> <S3: lm>
## 6 Bolivia (Plurinational State of) <tibble [34 x 3]> <S3: lm>
## 7 Brazil <tibble [34 x 3]> <S3: lm>
## 8 Canada <tibble [34 x 3]> <S3: lm>
## 9 Chile <tibble [34 x 3]> <S3: lm>
## 10 Colombia <tibble [34 x 3]> <S3: lm>
## # ... with 190 more rows, and 1 more variables: tidied <list># Add one more step that unnests the tidied column
country_coefficients <- nested %>%
mutate(model = map(data, ~ lm(percent_yes ~ year, data = .)),
tidied = map(model, tidy)) %>%
unnest(tidied)
# Print the resulting country_coefficients variable
head(country_coefficients)## # A tibble: 6 x 6
## country term estimate std.error statistic
## <chr> <chr> <dbl> <dbl> <dbl>
## 1 Afghanistan (Intercept) -11.063084650 1.4705189228 -7.523252
## 2 Afghanistan year 0.006009299 0.0007426499 8.091698
## 3 Argentina (Intercept) -9.464512565 2.1008982371 -4.504984
## 4 Argentina year 0.005148829 0.0010610076 4.852773
## 5 Australia (Intercept) -4.545492536 2.1479916283 -2.116159
## 6 Australia year 0.002567161 0.0010847910 2.366503
## # ... with 1 more variables: p.value <dbl># Print the country_coefficients dataset
head(country_coefficients)## # A tibble: 6 x 6
## country term estimate std.error statistic
## <chr> <chr> <dbl> <dbl> <dbl>
## 1 Afghanistan (Intercept) -11.063084650 1.4705189228 -7.523252
## 2 Afghanistan year 0.006009299 0.0007426499 8.091698
## 3 Argentina (Intercept) -9.464512565 2.1008982371 -4.504984
## 4 Argentina year 0.005148829 0.0010610076 4.852773
## 5 Australia (Intercept) -4.545492536 2.1479916283 -2.116159
## 6 Australia year 0.002567161 0.0010847910 2.366503
## # ... with 1 more variables: p.value <dbl># Filter for only the slope terms
country_coefficients %>% filter(term == "year")## # A tibble: 199 x 6
## country term estimate std.error
## <chr> <chr> <dbl> <dbl>
## 1 Afghanistan year 0.006009299 0.0007426499
## 2 Argentina year 0.005148829 0.0010610076
## 3 Australia year 0.002567161 0.0010847910
## 4 Belarus year 0.003907557 0.0007587624
## 5 Belgium year 0.003203234 0.0007652852
## 6 Bolivia (Plurinational State of) year 0.005802864 0.0009657515
## 7 Brazil year 0.006107151 0.0008167736
## 8 Canada year 0.001515867 0.0009552118
## 9 Chile year 0.006775560 0.0008220463
## 10 Colombia year 0.006157755 0.0009645084
## # ... with 189 more rows, and 2 more variables: statistic <dbl>,
## # p.value <dbl>Now let’s filter significant countries only by using our p-value strategy(p-value < 0.05)
# Filter for only the slope terms
slope_terms <- country_coefficients %>%
filter(term == "year")
# Add p.adjusted column, then filter
slope_terms %>% mutate(p.adjusted = p.adjust(p.value) )## # A tibble: 199 x 7
## country term estimate std.error
## <chr> <chr> <dbl> <dbl>
## 1 Afghanistan year 0.006009299 0.0007426499
## 2 Argentina year 0.005148829 0.0010610076
## 3 Australia year 0.002567161 0.0010847910
## 4 Belarus year 0.003907557 0.0007587624
## 5 Belgium year 0.003203234 0.0007652852
## 6 Bolivia (Plurinational State of) year 0.005802864 0.0009657515
## 7 Brazil year 0.006107151 0.0008167736
## 8 Canada year 0.001515867 0.0009552118
## 9 Chile year 0.006775560 0.0008220463
## 10 Colombia year 0.006157755 0.0009645084
## # ... with 189 more rows, and 3 more variables: statistic <dbl>,
## # p.value <dbl>, p.adjusted <dbl># Filter by adjusted p-values
filtered_countries <- country_coefficients %>%
filter(term == "year") %>%
mutate(p.adjusted = p.adjust(p.value)) %>%
filter(p.adjusted < .05)Now let’s look at the countries where the percentage of “Yes” votes increases the fastest.
# Sort for the countries increasing most quickly
filtered_countries %>% arrange(desc(estimate))## # A tibble: 61 x 7
## country term estimate std.error statistic
## <chr> <chr> <dbl> <dbl> <dbl>
## 1 South Africa year 0.011858333 0.0014003768 8.467959
## 2 Kazakhstan year 0.010955741 0.0019482401 5.623404
## 3 Yemen Arab Republic year 0.010854882 0.0015869058 6.840281
## 4 Kyrgyzstan year 0.009725462 0.0009884060 9.839541
## 5 Malawi year 0.009084873 0.0018111087 5.016194
## 6 Dominican Republic year 0.008055482 0.0009138578 8.814809
## 7 Portugal year 0.008020046 0.0017124482 4.683380
## 8 Honduras year 0.007717977 0.0009214260 8.376123
## 9 Peru year 0.007299813 0.0009764019 7.476238
## 10 Nicaragua year 0.007075848 0.0010716402 6.602820
## # ... with 51 more rows, and 2 more variables: p.value <dbl>,
## # p.adjusted <dbl>Now lets look at the countries where the percentage of “Yes” votes decreases the fastest
# Sort for the countries decreasing most quickly
filtered_countries %>% arrange(estimate)## # A tibble: 61 x 7
## country term estimate std.error statistic
## <chr> <chr> <dbl> <dbl> <dbl>
## 1 Republic of Korea year -0.009209912 0.0015453128 -5.959901
## 2 Israel year -0.006852921 0.0011718657 -5.847873
## 3 United States of America year -0.006239305 0.0009282243 -6.721764
## 4 Belgium year 0.003203234 0.0007652852 4.185673
## 5 Guinea year 0.003621508 0.0008326598 4.349325
## 6 Morocco year 0.003798641 0.0008603064 4.415451
## 7 Belarus year 0.003907557 0.0007587624 5.149908
## 8 Iran (Islamic Republic of) year 0.003911100 0.0008558952 4.569602
## 9 Congo year 0.003967778 0.0009220262 4.303324
## 10 Sudan year 0.003989394 0.0009613894 4.149613
## # ... with 51 more rows, and 2 more variables: p.value <dbl>,
## # p.adjusted <dbl>2- Conclusion
Using our simple linear regression model we were able to significantly observed voting trends during United Nations voting councils. We could further our analysis by merging new datasets with new information about specific factors that would push a country to vote yes and also make use of state-of-art machine learning algorithm for more accurate trend analysis.