Visualizing how tariff rates have changed around the world over the past few decades

tariff_viz.R

library(ggplot2)
library(reshape2)
library(dplyr)

# First, we need to get the data from the World Bank website.
# I am using the "weight mean of tariffs applied on manufactored
# products". The .csv file is below, but here is the url...
# https://data.worldbank.org/indicator/TM.TAX.MANF.WM.AR.ZS

wbt <- read.csv(file = 'C://your//file//path//wb_tariff.csv', header = T, stringsAsFactors = F)

# The file is a little messy, so we clean it up...
wbt <- wbt[,c(1,2,30:(ncol(wbt) -1))]   # Leave out years 1960 to 1985, due to lack of data points
wbt[,3:ncol(wbt)] <- lapply(wbt[,3:ncol(wbt)], function(x) as.numeric(x))      # Change tariff rates to numeric

wbt.melt <- melt(wbt)
wbt.melt$variable <- as.integer(substring(wbt.melt$variable, 2,5))    # Change 'year' column to integers
colnames(wbt.melt) <- c('country', 'code', 'year', 'tari')

# Nice, how our data is in a long format, which makes it
# easy to graph. We are going to use the geom_step function
# which (in my opinion) better illustrates the changing
# tariff rate in this case.
                    
myPlot <- ggplot(data = wbt.melt) +
  geom_step(aes(x=year, color=country, group = country, y=tari), na.rm = T) +
  theme(legend.position = 'none') +   # Since our data contains over 200 countries, we need to remove the legend!
  scale_y_continuous(trans = 'log', breaks = c(0.125,0.25,0.5,1,2,4,8,16,32,64,128), minor_breaks = NULL) +
  scale_x_continuous(breaks=c(1990,1995,2000,2005,2010,2015), minor_breaks = NULL)

# ~~~~ Take a look at the graph in the first comment below ~~~~
                            
# Next, let's add the worldwide average to this graph...
myPlot + geom_step(data = aggregate(by = list(wbt.melt$year), x = wbt.melt$tari, FUN = function(x) mean(x, na.rm = T)),
                  aes(x = Group.1, y = x), size = 2) +
  xlab('Year') + ylab('Average tariff %')       # Adding some axis labels

# The aggregate() function groups the data frame by the variable given in 'by',
# and applies the function given in 'FUN' to the values given in 'x'. In this
# case, we are telling aggregate() to group the <wbt.melt> data frame according
# to the $year column, and then find the average of $tari values for each year.
# The aggregate() function returns a data frame with columns $Group.1, $Group.2,
# $Group.3, etc, and an $x column. Because we only provided a single column to 
# group by (wbt.melt$year), the the $Group.1 column contains the years, and the
# $x column contains the mean value for each year.

# ~~~~ Take a look at the graph in the second comment below ~~~~

# Ok, this looks *pretty*, but it is not *informative*.
# It would be nice to see if there is a difference between the
# tariff rates in different parts of the world. The UN website
# provides a data we will need for this.
# Full .csv file is at end of page, and the url is below...
# https://unstats.un.org/unsd/methodology/m49/overview/

unRegion <- read.csv(file = 'C://Users//Sam//Documents//R//projects//tariff/UNSD.csv', header=T, stringsAsFactors = F)
names(unRegion)   # They provide a number of ____ for each country. We are going to use $Region.Name
unRegion <- unRegion[,c(4,9,11)]
colnames(unRegion) <- c('region', 'countryArea', 'code')
                   
# Ok, take a look at the <unRegion> data frame. You'll notice that
# each country has a 3-character country code which we will match
# up with country codes in our <wbt.melt> data frame.
head(unRegion)
                   
wbt.region <- merge(wbt.melt, unRegion, by = 'code', all = F)
# The merge() function is part of the dplyr package. It makes quick
# work of combining our very long <wbt.melt> and much shorter <unRegion>
# data frames.
# They are merged according to the $code column, which is common across
# both data frames.
# The 'all = F' argument tells it to leave out any rows which do not have
# a 3-character code in both $code columns. You will note that the merged
# data frame <wbt.region> is shorter than the <wbt.melt> data frame.
# It turns out that our original <wbt> data contained many "countries"
# which were actually "groups of countries". Thus, the merge() function
# left these rows out, because it did not find matching rows in the <unRegion>
# data frame.
                   
unique(wbt.melt$country) # See if you can find the ones which aren't countries...

head(wbt.region)    # Take a look at our new df. Let's make a new graph with the average for each region...

ggplot(data = wbt.region) +
  geom_step(aes(x=year, group= country, color = region, y=tari), alpha =0.4, na.rm = T) +
  # For the tiny step lines, we are grouping them by country, but coloring them according to their region
  geom_step(data =aggregate(x = wbt.region$tari, by = list(wbt.region$region, wbt.region$year), FUN = function(x) mean(x,na.rm = T)),
            aes(x=Group.2, y=x, color=Group.1), size=2) +
  # We are using the aggregate() function again to get the average tariff rate by year for each region.
  #
  xlab('Year') +
  ylab('Average tariff %') +
  guides(color = guide_legend(title = 'Region')) +
  scale_y_continuous(trans = 'log', breaks = c(1,2,4,8,0.5,0.25,0.125,16,32,64,128,256), minor_breaks = NULL) +
  scale_x_continuous(breaks=c(1990,1995,2000,2005,2010,2015), minor_breaks = c(1990:2017)) +
  ggtitle('Worldwide tariff rates',
          subtitle = 'Tiny lines are the rate for individual countries, colored by region.\nBold lines are region averages.') +
  theme(panel.grid.major.x = element_line(size = 1), plot.title = element_text())
  
  

world_tariff_applied_weighted_manf.csv

ww_country_codes.csv

The tariff rates of all the countries, with unique color for each country

The tariff rates for every country, with the worldwide mean

Finished plot!