Context
This work aims to replicate a graph presented by the Wall Street Journal (WSJ), in the year 2022. In that graph, data on Gross Domestic Product (GDP), in trillions and Research & Development expenditure (R&D) is presented for a set of countries in the year 2019. Those countries are then divided into two blocks, East and West, being the former build by data on the US, the EU, Japan and others, while the latter is conformed by China and Russia. Data is retrieved from the International Monetary Fund (IMF) database and the OECD database.
Figure 1: Original chart. Source: WSJ
Data loading and read in R
Data has been retrieved from the original sources the WSJ article exposed, thus being the IMF database for the GDP data and the OECD database for the R&D data. After major cleansing in Excel, data is exported in R as .csv, in order to perform minor changes and select countries to replicate the graph.
In what is related with OECD database, from where data on R&D has been recovered from the Main Science and Investigation Institute (MSTI)
In what is related with GDP data, recovered from the IMF database based on the World Economic Outlook (WEO). Both need to be loaded to be able to start.
# library (readr) or
gedr_data <- readr::read_delim("gedr_data.csv", delim = NULL,
show_col_types = FALSE)
gdp_data <- read.csv("gdp_data.csv", sep = ";")Even after some initial cleansing of the database in Excel, it is required to select the countries and columns of interest to graph. As long there are two graph that in the last step will be merged (one for GDP and other for R&D), I prefer to work with two separated databases rather than with one, in order to avoid misunderstandings.
library(tidyverse)
# 1st: I build a variable of countries selected. Note that China and China (
# People´s Republic of) refers to the same country, as they are coded different
# in both databases)
selected_countries <- c(
"China", "Russia", "North Korea", "United States", "Japan", "Canada",
"Australia","Austria", "Belgium", "Bulgaria", "Croatia", "Cyprus",
"Czech Republic", "Denmark","Estonia", "Finland", "France", "Germany",
"Greece", "Hungary", "Ireland", "Italy","Latvia", "Lithuania", "Luxembourg",
"Malta", "Netherlands", "Poland", "Portugal","Romania", "Slovak Republic",
"Slovenia", "Spain", "Sweden", "China (People’s Republic of)")
# 2nd: I filter gdp_data for selected countries and variables of interest,
# among other features. I use gdp_data as a base database, and gdp_data_ as a
# modified database.
gdp_data_ <- gdp_data |>
filter(Country %in% selected_countries,
stringr::str_detect(Subject.Descriptor, "current"),# for GDP data on current prices
stringr::str_detect(Units, "parity"),
stringr::str_detect(Scale, "Billions"))
# 3rd: Modify gdp_data_ to convert non-numeric columns into numeric cols.
# This is specially relevant for Date columns, that are displayed in XYEAR form
# and as chr.
gdp_data_ <- gdp_data_ |>
drop_na(all_of(paste0("X", 1990:2023))) |>
mutate(across(
X1990:X2023,
~ as.numeric(gsub(",", "", .x)) # remove commas, convert to numeric
)) |>
pivot_longer(
cols = X1990:X2023,
names_to = "year",
values_to = "gdp") |>
mutate(
year = as.integer(sub("X", "", year)))R&D database also needs to be cleaned, and only variables of interest selected.
# 1st: I clean and select countries in gedr_data. As in the previous, I use
# gedr_data as a base dataset, and a gedr_data_ as a modified one, in order to
# prevent misunderstandings and limit error losses.
gedr_data_ <- gedr_data |>
dplyr::select(`Reference area`, Measure, TIME_PERIOD, OBS_VALUE) |>
filter(
`Reference area` %in% selected_countries,
str_detect(Measure, "GERD"),
TIME_PERIOD == 2019)The first part of the cleaning and selection procedure is now done. However, as the original graph groups countries in blocks (EAST-WEST), and some other features (EU countries) can´t be extracted from the original databases, some work still needs to be done. Hence, some variables are created, such as eu_countries, west_allies, group or region.
# # 1st: I build a variable of eu_countries.
# Ensure names match your dataset (e.g., "Slovakia" vs "Slovak Republic")
eu_countries <- c(
"Austria","Belgium","Bulgaria","Croatia","Cyprus","Czech Republic","Denmark",
"Estonia","Finland","France","Germany","Greece","Hungary","Ireland","Italy",
"Latvia","Lithuania","Luxembourg","Malta","Netherlands","Poland","Portugal",
"Romania","Slovak Republic","Slovenia","Spain","Sweden")
# Explicitly defining Western allies to prevent them from being "lost"
west_allies <- c("United Kingdom", "Canada", "Australia", "Korea", "Israel", "Norway", "Switzerland")
# 1. Processing GDP Data. Add a group and region
gdp_data_ <- gdp_data_ |>
mutate(
group = case_when(
Country == "United States" ~ "U.S.",
Country == "Japan" ~ "Japan",
Country == "China" ~ "China",
Country == "Russia" ~ "Russia",
Country %in% eu_countries ~ "EU",
TRUE ~ "Other" # This captures allies and aligned countries
),
region = case_when(
# Define who belongs to the "WEST" side of the graph
group %in% c("U.S.", "Japan", "EU") | Country %in% west_allies ~ "WEST",
# Define who belongs to the "EAST" side (adjust list based on your data)
group %in% c("China", "Russia") ~ "EAST",
TRUE ~ "WEST"
)
) |>
# IMPORTANT: You must sum the values so "EU" is the total of all 27 countries
group_by(year, region, group) |>
summarise(gdp = sum(gdp, na.rm = TRUE), .groups = "drop")
# Replicates the process with GEDR
gedr_data_ <- gedr_data_ |>
mutate(
group = case_when(
`Reference area`== "United States" ~ "U.S.",
`Reference area`== "Japan" ~ "Japan",
`Reference area` == "China (People’s Republic of)" ~ "China",
`Reference area` == "Russia" ~ "Russia",
`Reference area`%in% eu_countries ~ "EU",
TRUE ~ "Other"),
region = case_when(
group %in% c("U.S.", "Japan", "EU", "Other") ~ "WEST",
group %in% c("China", "Russia") ~ "EAST"))Original Graph Replication
Now we are ready to start plotting our graph and try to replicate the original WSJ graph.
Firstly, we need to divide the whole graph into four different graphs, that in the end will be merged. That is to say, in the whole graph there are four different graphs, two corresponding to GDP and two corresponding to R&D. Concerning the GDP graphs, the blue one represents the GDP in trillions of the WEST block, that is to say the US, the EU, Japan, and other countries allied, such as Australia. On the other hand, the gold part of the graph replicates the GDP in trillions of the EAST block, meaning China and Russia, among others more difficult to know in base of the WSJ journal. It is needed to say that the gold part of the graph in GDP, is mirrored, so we need some prior transformations in our edited dataset (gdp_data_) to corectly plot it. The following chunk shows the mutation of the dataset to be able to plot the EAST block data.
gdp_data_$gdp_mirror <-
ifelse(gdp_data_$region == "EAST",
-gdp_data_$gdp,
gdp_data_$gdp)Now we are able to start plotting and replicating the graph. The following chunk starts to replicate the GDP part of the graph with the most basic version of our graph. Firstly, we will display our axis to then start plotting the data.
Once the code is run, we can see how our axis seem to be correct. Although they will be latter transformed, and labels year and gdp_mirror will be eliminated in future stages, before is needed to include some data.
Now we can see that our data is looking more similar to the graph we want to replicate. However, colors are not grouped by block, labels still there and the axis are not still the ones as in the WSJ article, so these changes are needed.
p <-
p + scale_x_continuous( # I create a vector of years to adequate it to WSJ.
breaks = c(1990, 1995, 2000, 2005, 2010, 2015, 2020),
labels = c("'90", "'95", "2000", "'05", "'10", "'15", "'20")) +
scale_y_continuous( # This vector is needed to not have negative numbers in EAST)
breaks = seq(-70000, 70000, 10000),
labels = c("$70","$60","50","40", "30", "20", "10", "0", "10", "20", "30", "$40",
"50", "60", "$70"),
position = "right",
expand = c(0,0))
print(p)
Now we can see how our graph is increasingly being more similar to the WSJ one. However, colors are still different, and not grouped by block. Moreover, labels are not present in the original WSJ graph, and some add no information, such as year.
p <-
p + scale_fill_manual(values = c(
"U.S." = "#94cced",
"EU" = "#94cced",
"Japan" = "#94cced",
"Other" = "#94cced",
"China" = "#e0c461",
"Russia" = "#e0c461")) +
labs(
title = "Gross domestic product, in trillions",
x = NULL,
y = NULL)
print(p)
After this step, the replication graph seems much more similar to the WSJ original. Now colors are the same, and axis replicate, within data, the ones of the original graph. In what is more, the title has been added, although it needs to be modified and placed in bold. Added to that, the group label is not present in the original one, and add few if any information. Besides, these labels will be added to be plotted within the graph, so there is no utility in keeping it.
p <-
p + theme_minimal(base_size = 11) +
theme(
panel.grid.minor = element_blank(),
panel.grid.major.x = element_blank(),
axis.title = element_blank(),
legend.position = "none",
plot.margin = margin(5, 10, 5, 5),
plot.title = element_text(size = 11, face = "bold")) +
annotate("text", x = 2010, y = 28000, label = "Other", size = 3.5, color = "black") +
annotate("text", x = 2010, y = 20000, label = "Japan", size = 3.5, color = "black") +
annotate("text", x = 2010, y = 16000, label = "EU", size = 3, color = "black") +
annotate("text", x = 2010, y = 8000, label = "U.S.", size = 3.5, color = "black") +
annotate("text", x = 2020, y = 5000, label = "WEST", size = 7, color = "white",
fontface = "bold", family = "Bale") +
annotate("text", x = 2010, y = -1500, label = "Russia", size = 3.5, color = "black") +
annotate("text", x = 2010, y = -10000, label = "China", size = 3.5, color = "black") +
annotate("text", x = 2020, y = -10000, label = "EAST", size = 7, color = "white",
fontface = "bold", family = "Bale")
print(p)
Now we can see we almost replicated the original graph. However, some major differences will be noted, such as the small portion of GDP the EU has. This is perhaps caused by differences in original dataset and our dataset. However, this seems a pretty accurate replication of our original graph, even taking into account these differences. By now, we will leave this graph stored as p, and start building the R&D graph, to then merge them with patchwork.
In what is related with R&D graph, two different graphs, one for WEST data and other for EAST data need to be built to then be merged. To make this process easier, I group gedr_data_ by group and region. This can be done in gedr_data_ but in order to not miss information I prefer to build a gedr_grouped and work from this results. In what is more, since two graphs are going to be done, I prefer to build a variable with colors rather than to repeat them twice.
gedr_grouped <- gedr_data_ |>
group_by(region, group) |>
summarise(gerd = sum(OBS_VALUE), .groups = "drop")
gedr_grouped$group <- factor(
gedr_grouped$group, levels = c("Other", "Japan", "EU", "U.S.", "Russia",
"China"))
fill_cols <- c(
"U.S." = "#94cced",
"EU" = "#94cced",
"Japan" = "#94cced",
"Other" = "#94cced",
"China" = "#e0c461",
"Russia" = "#e0c461"
)Now we can filter this data to build ordered different datasets for WEST and EAST, with data ordered from the lowest to the highest, and labels centered.
west_data <- filter(gedr_grouped, region == "WEST") |>
arrange(desc(group)) |> # to order data in descending order
mutate(
cumsum_gerd = cumsum(gerd), # to build the stacked bar
label_pos = cumsum_gerd - gerd/2 # to place the labels in the center.
)
east_data <- filter(gedr_grouped, region == "EAST") |>
arrange(desc(group)) |> # idem as before
mutate(
cumsum_gerd = cumsum(gerd), # idem as before
label_pos = cumsum_gerd - gerd/2 # idem as before
)Now we can start plotting each graph separately, starting from the WEST one.
p_west <-
ggplot(west_data, aes(x = 1, y = gerd, fill = group)) +
geom_bar(color = "white", width = 0.8, stat = "identity")
# stat = "identity" is used to display data on the dataset as it is.
print(p_west)
We can see that the initial steps of the process may lead us to replicate in a very precise manner the original WSJ graph. However, major changes need to be made, as including replicating colors, place names of the countries within the bars, eliminate the group label, place the y-axis in the right side or write the title.
p_west <- p_west +
geom_text(aes(y = label_pos, label = group),
color = "black", size = 3) +
scale_fill_manual(values = fill_cols) +
scale_y_continuous(
breaks = c(0, 10, 30, 50),
labels = c("", "10", "30", "$50"),
position = "right",
expand = c(0, 0)) +
scale_x_continuous(limits = c(0.5, 1.5)) +
labs(
title = "Gross domestic expenditure\non research and development\nin 2019, trillions"
) +
theme_minimal(base_size = 11) +
theme(
axis.title = element_blank(),
axis.text.x = element_blank(),
panel.grid = element_blank(),
legend.position = "none",
plot.title = element_text(size = 11, face = "bold", lineheight = 1.1),
plot.margin = margin(5, 5, 2, 10)
)
print(p_west)
Once done this, it seems that the resulting graph is a pretty accurate replication of the original WSJ we can repeat in the east part of the graph. In order to make it easier to read and faster to run, the following chunk imitates the process of the previous west graph at once. Note that, as they will be merged and the west part (the blue one) is placed at the top, when building the east part of the graph no title is needed.
p_east <- ggplot(east_data, aes(x = 1, y = gerd, fill = group)) +
geom_bar(color = "white", width = 0.8, stat = "identity") +
geom_text(aes(y = label_pos, label = group),
color = "black", size = 3) +
scale_y_continuous(
breaks = c(0, 1, 3),
labels = c("", "1", "$3"),
position = "right",
expand = c(0, 0)) +
scale_x_continuous(limits = c(0.5, 1.5)) +
scale_fill_manual(values = fill_cols) +
theme_minimal(base_size = 11) +
theme(
axis.title = element_blank(),
axis.text.x = element_blank(),
panel.grid = element_blank(),
legend.position = "none",
plot.margin = margin(2, 5, 5, 10)
)
print(p_east)
Now, we can merge them with patchwork to see the result when trying to replicate the whole right part of the WSJ journal. The following chunk libraries patchwork, that needs to be priorly installed, to then merge both of them. Note that the formula is used to place the east bars (the gold ones) below the west bars (the blue ones), as in the WSJ.
library(patchwork)
p_gedr <- p_west / p_east + plot_layout(heights = c(.77, .33))
print(p_gedr)
Once we have both parts of the original graph, being the GDP part that is placed in the left part of the graph and the R&D part that is placed in the right part of the graph, it is possible to merge them, add a general title and the notes placed in the original bottom part of the GDP part.
final_graph <- p + p_gedr +
plot_layout(widths = c(3, 1)) +
plot_annotation(
title = "The Western Advantage",
subtitle = "The U.S. and its democratic, market-based allies in Europe and Asia together generate far more economic\noutput and spend much more on research and development than China, Russia and countries aligned with\nthem.",
caption = "Note: GDP is converted from local currency to dollars at market exchange rates. Research and development spending is for\n2019, converted from local currency to dollars at purchasing power parity.\nSources: IMF (GDP); OECD (expenditure on R&D)",
theme = theme(
plot.title = element_text(size = 18, face = "bold"),
plot.subtitle = element_text(size = 12),
plot.caption = element_text(size = 10, hjust = 0, color = "gray40",
margin = margin(t = 15))
)
)
print(final_graph)
Limitations and furher improvement: It has been imposible, after numerous trials, to solve the mismatch between the original EU data and the replication EU data. This might have been provoked by differences in dataset exact sources, name mismatches not found when building eu_countries or code errors when building the graph.
Edited graph
Having done the previous, this graph can be redone in a cleaner, fancier and more understandable way. Firstly, the original graph have some perhaps minor areas of improvement, such as color differences. By plotting different variants of blue and gold, not only blocks but also countries are able to be differentiated. In what is more, the GDP type of graph correctly shows the overall evolution through almost two decades, but makes it difficult to see annual differences. Because of this, a new graph, with stacked bars plotted in form of pyramid, is built for the GDP data. In what is more, and for reader ease, the R&D data is displayed horizontally, to make it more alike to the GDP data. Once done this, both are merged with the same labels as the original one.
Said this, the first chunk prepares gdp_data_ to plot the edited version of this graph. Note that, as long now we have a mirrored graph, it is needed to build a negative (mirrored) variable for the EAST block, which is going to be in the left side of the graph. It can be done in only one line of code, but is preferred to do it separately for reader´s ease.
pyramid_gdp_data <- gdp_data_ |>
group_by(year, region, group) |>
summarise(gdp = sum(gdp, na.rm = TRUE), .groups = "drop")
# Create a mirrored data for EAST (build negative values on the left)
pyramid_gdp_data <- pyramid_gdp_data |>
mutate(
gdp_display = ifelse(region == "EAST", -gdp, gdp))Once done this, and in order to assure order is mantained, we need to order the column group in this new dataset pyramid_gdp_data. Moreover, since as said different tones of blue and gold would make the graph easier to read, the fill_cols variable needs to be redone, in order to save time at plotting colors. After different steps of trial and error, colors chosen are the ones displayed in the following.
pyramid_gdp_data$group <- factor(pyramid_gdp_data$group,
levels = c("Other", "Japan", "EU", "U.S.",
"Russia", "China"))
# Redefine colors
fill_cols <- c(
"U.S." = "#94cced",
"EU" = "#6baed6",
"Japan" = "#4292c6",
"Other" = "#2171b5",
"China" = "#e0c461",
"Russia" = "#d4a940")Now, all steps to be able to start plotting have been covered. The following chunk shows the inital steps for the pyramid GDP data to be plotted.
p_pyramid <- ggplot(pyramid_gdp_data, aes(x = gdp_display, y = as.factor(year),
fill = group)) +
geom_col(color = "white", linewidth = 0.3, position = "stack")
print(p_pyramid)
Now, we can see some initial intuition of the graph that is intended to improve the original WSJ one. However, some major changes need to be made, clean Y-axis and make the X-axistwo sided positive.
p_pyramid<- p_pyramid +
# X-axis cleansing
scale_x_continuous(
breaks = seq(-70000, 70000, 10000),
labels = c("$70", "60", "50", "40", "30", "20", "10", "0", "10", "20", "30",
"40", "50", "60", "$70"),
expand = c(0, 0),
limits = c(-72000, 72000)
) +
# Y-axis cleansing
scale_y_discrete(
breaks = as.character(seq(1990, 2023, 1)),
labels = c("1990", rep("", 4), "'95", rep("", 4), "'00", rep("", 4), "'05",
rep("", 4), "'10", rep("", 4),
"'15", rep("", 4), "'20", "", "", "'23"),
position = "right"
)
print(p_pyramid)
Moreover, and as said, new colors need to be plotted, axis labels can be eliminated and both region and country labels need to be added within the graph. Note that, for reader´ease country names have been included in the side of its block.
p_pyramid <- p_pyramid +
# Colors
scale_fill_manual(values = fill_cols) +
# Labels and annotations
labs(
title = "Gross Domestic Product by Year (1990-2023), in billions",
x = NULL,
y = NULL
) +
# Add region labels at the bottom (1990 level) in black
annotate("text", x = 35000, y = "1990", label = "WEST",
size = 8, color = "black", fontface = "bold") +
annotate("text", x = -35000, y = "1990", label = "EAST",
size = 8, color = "black", fontface = "bold") +
annotate("text", x = -65000, y = "2010", label = "China", size = 6,
color = "#e0c461", fontface = "bold") +
annotate("text", x = -65500, y = "2005", label = "Russia", size = 6,
color = "#d4a940", fontface = "bold") +
annotate("text", x = 65000, y = "2010", label = "US", size = 6,
color = "#94cced", fontface = "bold") +
annotate("text", x = 65000, y = "2005", label = "EU", size = 6,
color = "#6baed6", fontface = "bold") +
annotate("text", x = 65000, y = "2000", label = "Japan", size = 6,
color = "#4292c6", fontface = "bold") +
annotate("text", x = 65000, y = "1995", label = "Other", size = 6,
color = "#2171b5", fontface = "bold")
print (p_pyramid)
Now the graph seems much more WSJ stylish and clean. Lastly, it is time to add theme_minimal to make the graph even cleaner, remove group label on the side and finally add a black vertical line at 0 to make graph readability even more straightforward.
p_pyramid <- p_pyramid +
# theme_minimal and group label drop
theme_minimal(base_size = 15) +
theme(
panel.grid.minor = element_blank(),
panel.grid.major.y = element_blank(),
panel.grid.major.x = element_line(color = "gray85"),
legend.position = "none",
plot.title = element_text(size = 12, face = "bold"),
plot.subtitle = element_text(size = 12),
plot.margin = margin(10, 10, 10, 10),
axis.text = element_text(size = 8),
axis.text.y = element_text(hjust = 1)
) +
# Add a vertical line at zero
geom_vline(xintercept = 0, color = "black", linewidth = 0.8)
print(p_pyramid)
Regarding the R&D graph, little changes aside from color changes are thought to be needed. However, and to make readability more useful, the EAST gold part is now mirrored, with one black horizontal line built as a threshold between both blocks. The following chunk exposes the neccessary changes to mirror R&D graph for the EAST part, and the whole construction of the new R&D graph. As in the original p_gedr, the graph is divided in two parts (WEST and EAST), and then merged with patchwork.
gedr_grouped <- gedr_data_ |>
group_by(region, group) |>
summarise(gerd = sum(OBS_VALUE), .groups = "drop") #Perhaps already done, but
# to ensure it
gedr_grouped$group <- factor(
gedr_grouped$group, levels = c("Other", "Japan", "EU", "U.S.", "Russia",
"China"))
west_data <- filter(gedr_grouped, region == "WEST") %>%
arrange(desc(group)) %>%
mutate(
cumsum_gerd = cumsum(gerd),
label_pos = cumsum_gerd - gerd/2
)
p_west <- ggplot(west_data, aes(x = 1, y = gerd, fill = group)) +
geom_bar(color = "white", width = 0.8, stat = "identity") +
geom_text(aes(y = label_pos, label = group),
color = "black", size = 3) +
scale_fill_manual(values = fill_cols) +
scale_y_continuous(
breaks = c(0, 10, 30, 50),
labels = c("", "10", "30", "50"),
position = "right",
expand = c(0, 0)) +
scale_x_continuous(limits = c(0.5, 1.5)) +
labs(
title = "Gross domestic expenditure\non research and development\nin 2019, trillions"
) +
theme_minimal(base_size = 15) +
theme(
axis.title = element_blank(),
axis.text.x = element_blank(),
panel.grid = element_blank(),
legend.position = "none",
plot.title = element_text(size = 15, face = "bold", lineheight = 1.1),
plot.margin = margin(5, 5, 0, 10) # Changed bottom margin to 0
)
# --- EAST PANEL ---
east_data <- filter(gedr_grouped, region == "EAST") %>%
arrange(desc(group)) %>%
mutate(
cumsum_gerd = cumsum(gerd),
label_pos = cumsum_gerd - gerd/2,
gerd_neg = -gerd,
cumsum_neg = -cumsum_gerd,
label_pos_neg = -label_pos
)
p_east <- ggplot(east_data, aes(x = 1, y = gerd_neg, fill = group)) +
geom_bar(color = "white", width = 0.8, stat = "identity") +
geom_segment(aes(x = 0.6, xend = 1.4, y = 0, yend = 0),
color = "black", linewidth = 1) +
geom_text(aes(y = label_pos_neg, label = group),
color = "black", size = 3) +
scale_y_continuous(
breaks = c(-3, -1, 0),
labels = c("$3", "1", ""),
position = "right",
expand = c(0, 0)) +
scale_x_continuous(limits = c(0.5, 1.5)) +
scale_fill_manual(values = fill_cols) +
theme_minimal(base_size = 15) +
theme(
axis.title = element_blank(),
axis.text.x = element_blank(),
panel.grid = element_blank(),
legend.position = "none",
plot.margin = margin(0, 5, 5, 10)
)
# --- COMBINE WEST AND EAST---
p_gedr <- p_west / p_east + plot_layout(heights = c(.77, .33))
print(p_gedr)
Finally, both the R&D and the GDP (p_pyramid) parts are merged with patchwork, including titles and captions as in the original.
edited_graph <- p_pyramid + p_gedr +
plot_layout(widths = c(3, 1)) +
plot_annotation(
title = "The U.S. and its democratic, market-based allies in Europe and Asia together generate far more economic\noutput and spend much more on research and development than China, Russia and countries aligned with\nthem.",
caption = "Note: GDP is converted from local currency to dollars at market exchange rates. Research and development spending is for\n2019, converted from local currency to dollars at purchasing power parity.\nSources: IMF (GDP); OECD (expenditure on R&D)",
theme = theme(
plot.title = element_text(size = 15, margin = margin(b = 10)),
plot.caption = element_text(size = 10, hjust = 0, color = "gray40",
margin = margin(t = 10)))
)
print(edited_graph)
Final Graph
To ease graph readability and after some comments on the prior graph, some minor changes have been provided. Firstly, final datasets for both GEDR and GDP is done, to be able to build a scale to align both graphs.
# ------------------------------------------------------------------------------
# Dataset changes
# ------------------------------------------------------------------------------
# ------------------------------------------------------------------------------
# GDP Dataset modification to plot West in the left.
# ------------------------------------------------------------------------------
final_gdp_data <- gdp_data_ |>
group_by(year, region, group) |>
summarise(gdp = sum(gdp, na.rm = TRUE), .groups = "drop") |>
mutate(
# WEST negative → left
gdp_display = ifelse(region == "WEST", -gdp, gdp),
# Reverse stacking order ONLY for WEST
group_stack = if_else(
region == "WEST",
fct_rev(group),
group
)
)
# Factor levels
final_gdp_data$group_stack <- factor(
final_gdp_data$group_stack,
levels = c("China", "Russia", "U.S.", "EU", "Japan", "Other")
)
# ------------------------------------------------------------------------------
# GEDR dataset changes
# ------------------------------------------------------------------------------
# ------------------------------------------------------------------------------
# Dataset changes needed to plot it horizontally
# ------------------------------------------------------------------------------
final_gedr_data <- gedr_data_ |>
group_by(region, group) |>
summarise(gerd = sum(OBS_VALUE), .groups = "drop") |>
mutate(
gerd_display = ifelse(region == "WEST", -gerd, gerd)
)
final_gedr_data$group <- factor(
final_gedr_data$group,
levels = c("Other", "Japan", "EU", "U.S.", "Russia", "China")
)Once this is done, a scale factor is needed to try to align both graphs one above the another, and added to the GEDR final dataset, as is the one placed above.
# ------------------------------------------------------------------------------
# Find the maximum absolute values for proper scaling
# ------------------------------------------------------------------------------
max_gdp <- max(abs(final_gdp_data$gdp_display), na.rm = TRUE)
max_gerd <- max(abs(final_gedr_data$gerd_display), na.rm = TRUE)
# Calculate scaling factor to align R&D with GDP axis
scale_factor <- max_gdp / max_gerd
# Apply to final_gedr_data:
final_gedr_data <- final_gedr_data |>
mutate(
gerd_scaled = gerd_display * scale_factor,
year_label = "2019"
)That said,blocks WEST and EAST have been side-changed in the GDP graph, not only because of numbers provided but also by reader´s eye custom. Moreover, the R&D graph has been plotted horizontally at above the GDP graph, aligned at 0, to make easier the comparison between both graphs, blocks and countries. Said that, other minor changes such as recoloring when readability was not as favorable as required have been implemented.
# ------------------------------------------------------------------------------
# Graph
# ------------------------------------------------------------------------------
p_pyramid <- ggplot(
final_gdp_data,
aes(
x = gdp_display,
y = as.factor(year),
fill = group_stack
)
) +
geom_col(
color = "white",
linewidth = 0.3,
position = "stack"
) +
# X-axis
scale_x_continuous(
breaks = seq(-70000, 70000, 10000),
labels = c("$70", "60", "50", "40", "30", "20", "10", "0",
"10", "20", "30", "40", "50", "60", "$70"),
expand = c(0, 0),
limits = c(-72000, 72000)
) +
# Y-axis
scale_y_discrete(
breaks = as.character(seq(1990, 2023, 1)),
labels = c(
"1990", rep("", 4), "'95", rep("", 4), "'00", rep("", 4),
"'05", rep("", 4), "'10", rep("", 4),
"'15", rep("", 4), "'20", "", "", "'23"
),
position = "right"
) +
# Colors
scale_fill_manual(values = fill_cols) +
# Labels
labs(
title = "Gross Domestic Product by Year (1990–2023), in trillions",
x = NULL,
y = NULL
) +
# Region labels
annotate("text", x = -35000, y = "1993", label = "WEST",
size = 10, fontface = "bold") +
annotate("text", x = 35000, y = "1993", label = "EAST",
size = 10, fontface = "bold") +
# Country labels (unchanged logic)
annotate("text", x = 25000, y = "2010", label = "China",
size = 8, color = "#e0c461", fontface = "bold") +
annotate("text", x = 25000, y = "2005", label = "Russia",
size = 8, color = "#d4a940", fontface = "bold") +
annotate("text", x = -45000, y = "2010", label = "US",
size = 8, color = "#94cced", fontface = "bold") +
annotate("text", x = -45000, y = "2005", label = "EU",
size = 8, color = "#6baed6", fontface = "bold") +
annotate("text", x = -45000, y = "2000", label = "Japan",
size = 8, color = "#4292c6", fontface = "bold") +
annotate("text", x = -45000, y = "1995", label = "Other",
size = 8, color = "#2171b5", fontface = "bold") +
# Theme
theme_minimal(base_size = 20) +
theme(
panel.grid.minor = element_blank(),
panel.grid.major.y = element_blank(),
panel.grid.major.x = element_line(color = "gray85"),
legend.position = "none",
plot.title = element_text(size = 20, face = "bold"),
axis.text = element_text(size = 15),
axis.text.y = element_text(hjust = 1)
) +
# Zero line
geom_vline(xintercept = 0, linewidth = 0.8)
p_pyramid
Hence, the process is repeated with GERD graph, to plot it horizontally.
# ------------------------------------------------------------------------------
# West data and Graph
# ------------------------------------------------------------------------------
p_gerd <- ggplot(
final_gedr_data,
aes(x = gerd_scaled, y = year_label, fill = group)
) +
geom_col(
color = "white",
linewidth = 0.3,
position = position_stack(),
width = 0.6
) +
# Scale to match GDP visually, but with meaningful R&D labels
scale_x_continuous(
# Create breaks at clean intervals of 5 extending to 60
breaks = c(-55, -50, -45, -40, -35, -30, -25, -20, -15, -10, -5, 0, 5) * scale_factor,
labels = c("$55", "50", "45", "30", "35", "30", "25", "20", "15", "10", "5", "0", "$5"),
limits = c(-72000, 72000)
) +
scale_fill_manual(values = fill_cols) +
labs(
title = "Gross domestic expenditure on research and development, 2019 (billions)",
x = NULL,
y = NULL
) +
theme_minimal(base_size = 20) +
theme(
panel.grid = element_blank(),
axis.text.y = element_blank(),
axis.text.x = element_text(size = 15),
legend.position = "none",
plot.title = element_text(size = 20, face = "bold"),
plot.margin = margin(10, 10, 5, 10)
) +
# Zero line to match GDP graph
geom_vline(xintercept = 0, linewidth = 0.8)Lastly, both graphs are merged and aligned in this last step.
final_plot <- (p_gerd / p_pyramid) +
plot_layout(heights = c(1, 12)) + # Adjusted height ratio for better visibility
plot_annotation(
title = "The Western Advantage",
subtitle = "The U.S. and its democratic, market-based allies in Europe and Asia together generate far more economic\noutput and spend much more on research and development than China, Russia and countries aligned with\nthem.",
caption = "Note: GDP is converted from local currency to dollars at market exchange rates. Research and development spending is for\n2019, converted from local currency to dollars at purchasing power parity.\nSources: IMF (GDP); OECD (expenditure on R&D)",
theme = theme(
plot.title = element_text(size = 18, face = "bold"),
plot.subtitle = element_text(size = 12),
plot.caption = element_text(size = 10, hjust = 0, color = "gray40",
margin = margin(t = 15))
)
)
final_plot
Limitations and further improvement: Although both graphs are aligned at 0, because of the data nature numbers slightly differ (label $50 in the upper graph, and number 50 in the lower graph are not exactly aligned). Moreover, colors are flipped in the upper graph compared with the lower one. Because of the width of the bars, it has been decided not to plot any names, in order not to add redundant info. However, no midpoint is reached and when looking at it, the viewer might be confused at identifying which country expends most on research and development. Moreover, GERD is showed for year 2019, while GDP is available for year 2023, so there is a time-mismatch as there is no available data for GERD in 2023.
