Military spending in six major economies

2025

Reproducing and improving a chart on military spending from Our World in Data.

Author

Affiliation

Moritz Eckmann

 

Published

Jan. 6, 2026

Citation

Eckmann, 2026

Introduction

This project reproduces and extends a chart on military spending from Our World in Data (OWID). The original chart shows inflation–adjusted military expenditure for six countries (United States, Russia, China, India, United Kingdom, and France) from 1979 to 2023.

The goals of this project are:

Original chart

Original chart. Source: Barnum et al., Global Military Spending Dataset (2025), Our World in Data.

The original chart arranges the six countries in a 3x2 grid of small multiples. Each panel shows military spending over time, adjusted for inflation and expressed in US dollars. The design makes it relatively easy to see trends within each country, but it is harder to directly compare levels and growth rates across countries.

In the rest of the article I first reproduce this chart using the public dataset provided by OWID, and then propose an alternative visualization that makes cross–country comparisons more straightforward. Then, I propose two further possible alternatives.

Data and preparation

ms <- readr::read_csv("military-spending-gmsd.csv",
                      show_col_types = FALSE)

glimpse(ms)
Rows: 270
Columns: 6
$ Entity                                <chr> "China", "China", "Chi…
$ Code                                  <chr> "CHN", "CHN", "CHN", "…
$ Year                                  <dbl> 1979, 1980, 1981, 1982…
$ `Military expenditure (constant US$)` <dbl> 86628520000, 875424400…
$ `World regions according to OWID`     <chr> "Asia", "Asia", "Asia"…
$ time                                  <dbl> 2023, 2023, 2023, 2023…

The dataset contains one row per country and year, with military spending given in constant US dollars.

For this project I focus on the six countries that appear in the original chart and convert spending to billions of US dollars to match the scale in the figure.

countries_focus <- c(
  "United States", "Russia", "China",
  "India", "United Kingdom", "France"
)

ms_clean <- ms %>%
  filter(Entity %in% countries_focus) %>%
  mutate(
    country = factor(
      Entity,
      levels = c(
        "United States",
        "Russia",
        "China",
        "India",
        "United Kingdom",
        "France"
      )
    ),
    spending_billion = `Military expenditure (constant US$)` / 1e9
  )

# Helper for y-axis labels: 200 -> $200B, 1200 -> $1.2T
format_spend <- function(x) {
  ifelse(
    x >= 1000,
    paste0("$", x / 1000, "T"),
    paste0("$", x, "B")
  )
}

Replicating the original small–multiples chart

In this section I recreate the original 3×2 grid of country panels.
The goal is to match the Our World in Data chart as closely as possible, using a helper function for a single country and then combining six panels into a grid.

Step 1: Define the helper function:

# White background for the replicated chart
bg_rep <- "white"

# Helper function: create the panel for a single country
make_country_plot <- function(ctry) {
  ggplot(
    filter(ms_clean, country == ctry),
    aes(x = Year, y = spending_billion)
  ) +
    geom_line(color = "steelblue", linewidth = 0.8) +
    scale_x_continuous(
      breaks = c(1979, 1990, 2000, 2010, 2023),
      limits = c(1979, 2023),
      expand = expansion(mult = c(0, 0))
    ) +
    scale_y_continuous(
      limits = c(0, NA),
      labels = format_spend,
      expand = expansion(mult = c(0, 0.05))
    ) +
    labs(
      tag = ctry,
      x   = NULL,
      y   = NULL
    ) +
    theme_minimal(base_size = 12) +
    theme(
      panel.background   = element_rect(fill = bg_rep, colour = NA),
      plot.background    = element_rect(fill = bg_rep, colour = NA),
      panel.grid.major.x = element_blank(),
      panel.grid.minor.x = element_blank(),
      panel.grid.major.y = element_line(
        linetype = "dashed",
        colour   = "grey80"
      ),
      panel.grid.minor.y = element_blank(),
      axis.text.x        = element_text(colour = "grey30"),
      axis.text.y        = element_text(colour = "grey30"),
      # move country label up and slightly right
      plot.tag           = element_text(
        hjust   = 0,
        face    = "bold",
        size    = 12,
        family = "sans"
      ),
      plot.tag.position  = c(0.03, 1.10),
      plot.margin        = margin(t = 30, r = 5, b = 10, l = 45)
    )
}

Step 2: Create one panel per country

Here I call make_country_plot() once for each country in the original figure. This reuses exactly the same plotting logic six times, which keeps the code short and ensures that all panels share identical scales and styling.

p_us <- make_country_plot("United States")
p_ru <- make_country_plot("Russia")
p_cn <- make_country_plot("China")
p_in <- make_country_plot("India")
p_uk <- make_country_plot("United Kingdom")
p_fr <- make_country_plot("France")

Step 3: Arrange panels in a 3×2 grid

In this chunk I combine the six panels into a 3×2 grid using the patchwork operators:

The result is the same small–multiples layout as in the Our World in Data chart.

p_countries <- (p_us | p_ru) /
  (p_cn | p_in) /
  (p_uk | p_fr)

p_countries

Step 4: Add shared title and subtitle

Finally, I add a shared title and subtitle using plot_annotation(). The additional theme() call customises the title and subtitle appearance and makes sure the overall background is also white. The last line prints p_replication, so the completed replication plot appears in the knitted document.

p_replication <- p_countries +
  plot_annotation(
    title    = "Military spending",
    subtitle = "This data is expressed in US dollars and adjusted for inflation."
  ) &
  theme(
    plot.background = element_rect(fill = bg_rep, colour = NA),
    plot.title = element_text(
      hjust  = 0,
      colour = "grey20",
      face   = "bold",
      size   = 20,
      family = "serif"
    ),
    plot.subtitle = element_text(
      hjust  = 0,
      colour = "grey40",
      size   = 12,
      family = "sans"
    )
  )

p_replication

Improved visualization: all countries in one chart

The small–multiples design makes it easy to focus on one country at a time, but it is harder to compare levels and growth rates across countries. As an alternative, I plot all six countries in a single chart and enrich the context with a few key historical events.

Step 1: Palette and overall y-axis

In this first chunk I define:

# Colorblind-friendly palette (Okabe–Ito)
okabe_ito <- c(
  "United States"   = "#E69F00",
  "Russia"          = "#56B4E9",
  "China"           = "#009E73",
  "India"           = "#CC79A7",
  "United Kingdom"  = "#0072B2",
  "France"          = "#D55E00"
)

# Maximum y value (rounded) for the plot
y_max <- ceiling(
  max(ms_clean$spending_billion, na.rm = TRUE) / 100
) * 100

Step 2: Line-end labels

This chunk prepares the labels that appear at the end of each line:

The resulting data frame labels_df is later used in geom_text() to write country names directly next to the lines instead of using a legend.

labels_raw <- ms_clean %>%
  dplyr::group_by(country) %>%
  dplyr::filter(Year == max(Year)) %>%
  dplyr::ungroup()

small_countries <- c("France", "United Kingdom", "India", "Russia")

min_small <- min(
  labels_raw$spending_billion[labels_raw$country %in% small_countries]
) - 20
step_small <- 28

labels_df <- labels_raw %>%
  dplyr::mutate(
    label_text = dplyr::if_else(
      country == "United Kingdom",
      "UK",
      as.character(country)
    ),
    label_y = dplyr::case_when(
      country == "France"         ~ min_small + 0 * step_small,
      country == "United Kingdom" ~ min_small + 1 * step_small,
      country == "India"          ~ min_small + 2 * step_small,
      country == "Russia"         ~ min_small + 3 * step_small,
      TRUE                        ~ spending_billion
    )
  )

Step 3: Annotating key historical events

Here I define events_df, a small tibble with:

These events provide context and help explain some of the changes in military spending over time.

events_df <- tibble::tibble(
  year  = c(1990, 2001, 2017),
  label = c(
    "End of Cold War",
    "9/11 attacks",
    "Trump 1st presidency"
  )
)

Step 4: Putting everything together in one chart

The final chunk builds the actual chart:

y_lim <- y_max * 1.08

ggplot(
  ms_clean,
  aes(x = Year, y = spending_billion, colour = country)
) +
  geom_line(linewidth = 0.9) +
  geom_segment(
    data = events_df,
    aes(
      x = year, xend = year,
      y = 0,    yend = y_lim * 0.96
    ),
    inherit.aes = FALSE,
    linetype = "dashed",
    colour = "grey60"
  ) +
  geom_text(
    data = events_df,
    aes(
      x = year,
      y = y_lim * 0.985,
      label = label
    ),
    inherit.aes = FALSE,
    vjust = 0, 
    size = 3.4,
    family = "sans",
    fontface = "bold",
    colour = "grey40"
  ) +
  geom_text(
    data = labels_df,
    aes(
      x = Year,
      y = label_y,
      label = label_text
    ),
    hjust = 0,
    nudge_x = 0.5,
    size = 3.4,
    show.legend = FALSE
  ) +
  scale_x_continuous(
    breaks = c(1979, 1990, 2000, 2010, 2023),
    limits = c(1979, 2025),
    expand = expansion(mult = c(0, 0))
  ) +
  scale_y_continuous(
    limits = c(0, y_lim),
    breaks = seq(0, y_max, by = 200),
    labels = format_spend,
    expand = expansion(mult = c(0, 0.05))
  ) +
  scale_colour_manual(values = okabe_ito, name = NULL) +
  coord_cartesian(clip = "off") +
  labs(
    title = "Military spending",
    subtitle = paste(
      "Six countries, constant 2022 US dollars.",
      "Lines show annual military expenditure;",
      "vertical dashed lines mark key geopolitical events."
    ),
    x = NULL,
    y = NULL
  ) +
  theme_minimal(base_size = 12) +
  theme(
    panel.background   = element_rect(fill = bg_col, colour = NA),
    plot.background    = element_rect(fill = bg_col, colour = NA),
    panel.grid.major.x = element_blank(),
    panel.grid.minor.x = element_blank(),
    panel.grid.major.y = element_line(
      linetype = "dashed",
      colour  = "grey80"
    ),
    panel.grid.minor.y = element_blank(),
    axis.text          = element_text(colour = "grey30"),
    legend.position    = "none",
    plot.margin        = margin(t = 15, r = 60, b = 10, l = 10),
    plot.title = element_text(
      hjust  = 0,
      colour = "grey20",
      face   = "bold",
      size   = 18,
      family = "serif"
    ),
    plot.subtitle = element_text(
      hjust  = 0,
      colour = "grey40",
      size   = 11,
      family = "sans"
    )
  )

Because all countries share the same axes, it becomes easier to compare both absolute levels and trends across the six countries than in the original small-multiples design.

Alternatives

The Graph still has weaknesses. Firstly, it gets difficult to read the development from Russia, India, UK and France. So an alternative to the absolut y-axis might make sense. Secondly, it fails to depict the relative development. Therefore, two alternatives are introduced.

Alternative 1: Logarithmic y-axis

The absolute plot with all six countries in one panel makes it easy to compare spending levels, but the United States dominates the scale. Spending in China and India is much smaller in absolute terms, even though their growth rates are very large.

A common way to emphasise proportional changes is to use a logarithmic y-axis. On a log scale, equal vertical distances correspond to equal percentage changes. This makes it easier to compare growth patterns across countries, at the cost of making absolute differences in levels less intuitive.

min_pos <- min(
  ms_clean$spending_billion[ms_clean$spending_billion > 0],
  na.rm = TRUE
)

y_lim_log <- y_max * 1.20

log_breaks <- c(10, 20, 50, 100, 200, 500, 1000)
log_breaks <- log_breaks[log_breaks >= min_pos & log_breaks <= y_lim_log]

ggplot(
  ms_clean,
  aes(x = Year, y = spending_billion, colour = country)
) +
  geom_line(linewidth = 0.9) +
  geom_segment(
    data = events_df,
    aes(
      x = year, xend = year,
      y = min_pos, yend = y_lim_log * 0.96
    ),
    inherit.aes = FALSE,
    linetype = "dashed",
    colour = "grey60"
  ) +
  geom_text(
    data = events_df,
    aes(
      x = year,
      y = y_lim_log * 0.985,
      label = label
    ),
    inherit.aes = FALSE,
    vjust = 0,
    size = 3.4,
    family = "sans",
    fontface = "bold",
    colour = "grey40"
  ) +
  geom_text(
    data = labels_df,
    aes(x = Year, y = label_y, label = label_text),
    hjust = 0,
    nudge_x = 0.5,
    size = 3.4,
    show.legend = FALSE
  ) +
  scale_x_continuous(
    breaks  = c(1979, 1990, 2000, 2010, 2023),
    limits  = c(1979, 2025),
    expand  = expansion(mult = c(0, 0))
  ) +
  scale_y_log10(
    limits = c(min_pos, y_lim_log),
    breaks = log_breaks,
    labels = format_spend,
    expand = expansion(mult = c(0, 0.05))
  ) +
  scale_colour_manual(values = okabe_ito, name = NULL) +
  coord_cartesian(clip = "off") +
  labs(
    title = "Military spending (log scale)",
    subtitle = paste(
      "Same axis scale for all six countries,",
      "logarithmic y-axis (constant 2022 US dollars)."
    ),
    x = NULL,
    y = NULL
  ) +
  theme_minimal(base_size = 12) +
  theme(
    panel.background   = element_rect(fill = bg_col, colour = NA),
    plot.background    = element_rect(fill = bg_col, colour = NA),
    panel.grid.major.x = element_blank(),
    panel.grid.minor.x = element_blank(),
    panel.grid.major.y = element_line(
      linetype = "dashed",
      colour  = "grey80"
    ),
    panel.grid.minor.y = element_blank(),
    axis.text          = element_text(colour = "grey30"),
    legend.position    = "none",
    plot.margin        = margin(t = 15, r = 60, b = 10, l = 10),
    plot.title = element_text(
      hjust  = 0,
      colour = "grey20",
      face   = "bold",
      size   = 18,
      family = "serif"
    ),
    plot.subtitle = element_text(
      hjust  = 0,
      colour = "grey40",
      size   = 11,
      family = "sans"
    )
  )

The log-scale version is less insightful because the non-linear y-axis is harder to interpret and visually compresses differences between countries. It highlights proportional growth, but the distorted axis can easily confuse readers who want to compare actual spending levels.

Alternative 2: Relative change since 1979 (index plot)

Another way to emphasise relative change is to normalise each country’s spending to a common starting value. Here I set spending in 1979 to 100 for each country and plot an index over time. A value of 200 means that spending has doubled relative to 1979.

This design deliberately hides information about absolute levels and focuses instead on the question: who has grown the most since 1979? It highlights that countries like China and India have experienced much faster proportional increases than more established military powers.

ms_index <- ms_clean %>%
  dplyr::filter(!is.na(spending_billion)) %>%
  dplyr::group_by(country) %>%
  dplyr::mutate(
    base_1979 = spending_billion[Year == 1979][1]
  ) %>%
  dplyr::filter(!is.na(base_1979) & base_1979 > 0) %>%
  dplyr::mutate(
    spending_index = spending_billion / base_1979 * 100
  ) %>%
  dplyr::ungroup() %>%
  dplyr::filter(!is.na(spending_index))

y_max_idx <- ceiling(
  max(ms_index$spending_index, na.rm = TRUE) / 50
) * 50

y_lim_idx <- y_max_idx * 1.08

labels_idx <- ms_index %>%
  dplyr::group_by(country) %>%
  dplyr::filter(Year == max(Year)) %>%
  dplyr::ungroup() %>%
  dplyr::mutate(
    label_text = dplyr::if_else(
      country == "United Kingdom",
      "UK",
      as.character(country)
    )
  )

events_idx <- tibble::tibble(
  year  = c(1990, 2001, 2017),
  label = c(
    "End of Cold War",
    "9/11 attacks",
    "Trump 1st presidency"
  )
)

ggplot(
  ms_index,
  aes(x = Year, y = spending_index, colour = country)
) +
  geom_line(linewidth = 0.9) +
  geom_segment(
    data = events_idx,
    aes(
      x = year, xend = year,
      y = 0,    yend = y_lim_idx * 0.96
    ),
    inherit.aes = FALSE,
    linetype = "dashed",
    colour = "grey60"
  ) +
  geom_text(
    data = events_idx,
    aes(
      x = year,
      y = y_lim_idx * 0.985,
      label = label
    ),
    inherit.aes = FALSE,
    vjust = 0,
    size = 3.4,
    family = "sans",
    fontface = "bold",
    colour = "grey40"
  ) +
  geom_text(
    data = labels_idx,
    aes(
      x = Year,
      y = spending_index,
      label = label_text
    ),
    hjust = 0,
    nudge_x = 0.5,
    size = 3.4,
    show.legend = FALSE
  ) +
  scale_x_continuous(
    breaks  = c(1979, 1990, 2000, 2010, 2023),
    limits  = c(1979, 2025),
    expand  = expansion(mult = c(0, 0))
  ) +
  scale_y_continuous(
    limits  = c(0, y_lim_idx),
    breaks  = seq(0, y_max_idx, by = 50),
    labels  = function(x) paste0(round(x), " (1979 = 100)"),
    expand  = expansion(mult = c(0, 0.05))
  ) +
  scale_colour_manual(values = okabe_ito, name = NULL) +
  coord_cartesian(clip = "off") +
  labs(
    title = "Military spending indexed to 1979 = 100",
    subtitle = paste(
      "Relative change in military spending.",
      "Each country starts at 100 in 1979."
    ),
    x = NULL,
    y = NULL
  ) +
  theme_minimal(base_size = 12) +
  theme(
    panel.background   = element_rect(fill = bg_col, colour = NA),
    plot.background    = element_rect(fill = bg_col, colour = NA),
    panel.grid.major.x = element_blank(),
    panel.grid.minor.x = element_blank(),
    panel.grid.major.y = element_line(
      linetype = "dashed",
      colour  = "grey80"
    ),
    panel.grid.minor.y = element_blank(),
    axis.text          = element_text(colour = "grey30"),
    legend.position    = "none",
    plot.margin        = margin(t = 15, r = 60, b = 10, l = 10),
    plot.title = element_text(
      hjust  = 0,
      colour = "grey20",
      face   = "bold",
      size   = 18,
      family = "serif"
    ),
    plot.subtitle = element_text(
      hjust  = 0,
      colour = "grey40",
      size   = 11,
      family = "sans"
    )
  )

The indexed chart clearly shows relative growth, but it hides all information about the actual size of military budgets. This makes it good for ranking growth rates, but less suitable than the main improved chart if we care about both growth and the absolute scale of spending.

Conclusions and further work

Replicating the original chart shows that small multiples work well for seeing each country separately, but make cross-country comparisons harder. The improved combined chart, with a shared scale, line-end labels and event annotations, makes it much easier to compare both levels and trends across all six countries at once.

The log-scale and indexed alternatives are useful for exploring specific questions about proportional growth, but their axes are less intuitive and hide absolute spending levels, so they work better as complementary views than as the main figure. As next steps, one could normalise spending by GDP or population, add more countries or alliances, or build an interactive version that lets readers switch between these different perspectives.

Footnotes

    Reuse

    Text and figures are licensed under Creative Commons Attribution CC BY 4.0. The figures that have been reused from other sources don't fall under this license and can be recognized by a note in their caption: "Figure from ...".

    Citation

    For attribution, please cite this work as

    Eckmann (2026, Jan. 7). Data Visualization | MSc CSS: Military spending in six major economies. Retrieved from https://csslab.uc3m.es/dataviz/projects/2025/100572935/

    BibTeX citation

    @misc{eckmann2026military,
  author = {Eckmann, Moritz},
  title = {Data Visualization | MSc CSS: Military spending in six major economies},
  url = {https://csslab.uc3m.es/dataviz/projects/2025/100572935/},
  year = {2026}
}