Code
library(tidyverse)Project 2 — Data Tidying: GDP per Capita
Introduction
This is one of the three datasets for Project 2. In this file, I will tidy my own GDP per capita dataset, which I shared in the 5A discussion posts. I will decide the tidying steps myself, based on what makes the data messy, and then use the analysis ideas I suggested in my post.
Source and Requested Analysis
This dataset is the GDP per capita (current US$) data from the World Bank (https://data.worldbank.org/indicator/NY.GDP.PCAP.CD), which I downloaded as a CSV file. It shows GDP per capita for countries and regions from 1960 to 2025. I suggested comparing how GDP per capita has changed over time in different countries and regions and finding which countries had the fastest growth in recent years.
Data Structure
The data is wide because each year from 1960 to 2025 has its own column. In the tidy version, each row will show one country in one year, with country, year, and GDP per capita.
Planned Approach
First, I will upload the World Bank CSV file to GitHub as it is, in its original wide format, and read it into R from there with skip = 4. Next, I will remove the empty column at the end, move the rows for regions and groups into a separate table, and use one pivot_longer() to put all the years into a single year column. I will also rename the columns in one consistent style and change the year to a number.
Once the data is tidy, I will compare how GDP per capita changed over time in different countries, calculate which countries grew the fastest in recent years, and compare a few regions and income groups. I will use summary tables and ggplot2 charts with clear titles and labels, and end with a conclusion that refers to specific values.
Anticipated Challenges
The file has metadata lines at the top, so reading it without skip = 4 gives an error, and there is an empty column at the end. The rows mix countries with regions and groups like “Euro area” or “World”, so I will move these into a separate table and keep only countries in the main table. Many countries also have missing values in the early years; for example, Aruba has no values before 1986. I will keep these as NA instead of replacing them with 0, and to compare growth fairly, I will use a time range where most countries have data, even though this means leaving out some early years.
Tidying the Data
Loading the packages
Reading CSV
Code
gdp_raw <- read.csv(
"https://raw.githubusercontent.com/esradogan3/Project_2/refs/heads/main/API_NY.GDP.PCAP.CD_DS2_en_csv_v2_429627.csv",
skip = 4
)
# Structure before tidying
dim(gdp_raw) # rows and columns[1] 265 71Code
names(gdp_raw)[1:8] # first column names[1] "Country.Name" "Country.Code" "Indicator.Name" "Indicator.Code"
[5] "X1960" "X1961" "X1962" "X1963" Code
tail(names(gdp_raw), 3) # last column names[1] "X2024" "X2025" "X" Code
head(gdp_raw[, 1:8]) # representative sample Country.Name Country.Code Indicator.Name
1 Aruba ABW GDP per capita (current US$)
2 Africa Eastern and Southern AFE GDP per capita (current US$)
3 Afghanistan AFG GDP per capita (current US$)
4 Africa Western and Central AFW GDP per capita (current US$)
5 Angola AGO GDP per capita (current US$)
6 Albania ALB GDP per capita (current US$)
Indicator.Code X1960 X1961 X1962 X1963
1 NY.GDP.PCAP.CD NA NA NA NA
2 NY.GDP.PCAP.CD 186.0895 186.9094 197.3679 225.4005
3 NY.GDP.PCAP.CD NA NA NA NA
4 NY.GDP.PCAP.CD 121.9368 127.4510 133.8238 139.0050
5 NY.GDP.PCAP.CD NA NA NA NA
6 NY.GDP.PCAP.CD NA NA NA NAI read the raw CSV file from my GitHub repository. The first four lines of the file are metadata (the data source and the update date), so I used skip = 4 to start reading at the real header. The raw data has 265 rows and 71 columns. Because the year columns start with numbers, read.csv() added an “X” in front of them (X1960, X1961, and so on), and the empty column at the end of every line became a column called X.
Separating regions and groups
The World Bank file mixes 217 countries with 48 regions and groups, such as “World”, “Euro area”, and “High income”. These groups are sums or averages of countries, so if I kept them together with the countries, the same people would be counted twice. Instead of deleting them, I will move them into a separate table so I can still compare them. To make this list, I used an LLM (Claude) to go through all 265 names in the file and pick the ones that are regions or groups instead of countries. I then checked the list myself: it has 48 codes, and the remaining 217 rows are all countries.
Code
aggregate_codes <- c(
"AFE", "AFW", "ARB", "CEB", "CSS", "EAP", "EAR", "EAS", "ECA", "ECS",
"EMU", "EUU", "HIC", "HPC", "IBD", "IBT", "IDA", "IDB", "IDX", "INX",
"LAC", "LCN", "LDC", "LIC", "LMC", "LMY", "LTE", "MEA", "MIC", "MNA",
"NAC", "OED", "OSS", "PRE", "PSS", "PST", "SAS", "SSA", "SSF", "SST",
"TEA", "TEC", "TLA", "TMN", "TSA", "TSS", "UMC", "WLD"
)
# Check: these are the rows I will move to the regions table
gdp_raw |>
filter(Country.Code %in% aggregate_codes) |>
select(Country.Code, Country.Name) |>
head(10) Country.Code Country.Name
1 AFE Africa Eastern and Southern
2 AFW Africa Western and Central
3 ARB Arab World
4 CEB Central Europe and the Baltics
5 CSS Caribbean small states
6 EAP East Asia & Pacific (excluding high income)
7 EAR Early-demographic dividend
8 EAS East Asia & Pacific
9 ECA Europe & Central Asia (excluding high income)
10 ECS Europe & Central AsiaTidy Format
Code
gdp_long <- gdp_raw |>
# Remove the empty last column and the two columns that are the same in every row
select(-X, -Indicator.Name, -Indicator.Code) |>
# Rename the columns in one consistent lowercase style
rename(
country = Country.Name,
country_code = Country.Code
) |>
# Put all the year columns into one year column
pivot_longer(
cols = starts_with("X"),
names_to = "year",
names_prefix = "X",
values_to = "gdp_per_capita"
) |>
# Change year to a number
mutate(year = as.integer(year))
# Split into countries and regions/groups
gdp_tidy <- gdp_long |>
filter(!country_code %in% aggregate_codes)
gdp_regions <- gdp_long |>
filter(country_code %in% aggregate_codes) |>
rename(region = country, region_code = country_code)
gdp_tidy# A tibble: 14,322 × 4
country country_code year gdp_per_capita
<chr> <chr> <int> <dbl>
1 Aruba ABW 1960 NA
2 Aruba ABW 1961 NA
3 Aruba ABW 1962 NA
4 Aruba ABW 1963 NA
5 Aruba ABW 1964 NA
6 Aruba ABW 1965 NA
7 Aruba ABW 1966 NA
8 Aruba ABW 1967 NA
9 Aruba ABW 1968 NA
10 Aruba ABW 1969 NA
# ℹ 14,312 more rowsCode
gdp_regions# A tibble: 3,168 × 4
region region_code year gdp_per_capita
<chr> <chr> <int> <dbl>
1 Africa Eastern and Southern AFE 1960 186.
2 Africa Eastern and Southern AFE 1961 187.
3 Africa Eastern and Southern AFE 1962 197.
4 Africa Eastern and Southern AFE 1963 225.
5 Africa Eastern and Southern AFE 1964 209.
6 Africa Eastern and Southern AFE 1965 227.
7 Africa Eastern and Southern AFE 1966 241.
8 Africa Eastern and Southern AFE 1967 244.
9 Africa Eastern and Southern AFE 1968 257.
10 Africa Eastern and Southern AFE 1969 282.
# ℹ 3,158 more rowsI removed the empty column at the end, and also the Indicator.Name and Indicator.Code columns, because they have the same value (“GDP per capita (current US$)”) in every row. I renamed the columns to country and country_code so that all names follow the same lowercase style, and moved the 48 regions and groups into a separate table called gdp_regions. Since each year column holds only one piece of information, the year, one pivot_longer() was enough. I used names_prefix = “X” to remove the “X” that read.csv() added, and then changed year to an integer.
Checking missing and inconsistent values
Code
# Size of the tidy data: 217 countries x 66 years
n_distinct(gdp_tidy$country)[1] 217Code
nrow(gdp_tidy)[1] 14322Code
# Total number of missing values
sum(is.na(gdp_tidy$gdp_per_capita))[1] 2558Code
# Number of countries with data in selected years
gdp_tidy |>
filter(year %in% c(1960, 1990, 2015, 2024, 2025)) |>
group_by(year) |>
summarise(countries_with_data = sum(!is.na(gdp_per_capita)))# A tibble: 5 × 2
year countries_with_data
<int> <int>
1 1960 111
2 1990 192
3 2015 212
4 2024 200
5 2025 186Code
# Example: the first year with data for Aruba
gdp_tidy |>
filter(country == "Aruba", !is.na(gdp_per_capita)) |>
slice_min(year)# A tibble: 1 × 4
country country_code year gdp_per_capita
<chr> <chr> <int> <dbl>
1 Aruba ABW 1986 6768.There are 2,558 missing values out of 14,322 rows. Most of them are in the early years: only 111 countries have data in 1960, but 212 have data in 2015. Aruba, for example, has no values before 1986. I kept these as NA instead of replacing them with 0, because 0 would mean a country had no income at all, which is not true. The year 2025 also has fewer countries (186) than 2024 (200), because some countries have not reported yet.
Code
# One country name is written in a different language
gdp_tidy |>
filter(country_code == "NRU") |>
distinct(country, country_code)# A tibble: 1 × 2
country country_code
<chr> <chr>
1 Naoero NRU Code
gdp_tidy <- gdp_tidy |>
mutate(country = recode(country, "Naoero" = "Nauru"))I also found one inconsistent country name: Nauru is written as “Naoero”, its name in the Nauruan language, while all other countries use English names. I changed it to “Nauru”.
Analysis
GDP per capita over time in selected countries
Code
selected <- c("United States", "Germany", "Japan", "Turkiye", "China", "Brazil", "India")
gdp_tidy |>
filter(country %in% selected, !is.na(gdp_per_capita)) |>
ggplot(aes(x = year, y = gdp_per_capita, color = country)) +
geom_line() +
scale_y_log10(labels = scales::comma) +
labs(
title = "GDP per Capita in Selected Countries (1960–2025)",
x = "Year",
y = "GDP per capita (current US$, log scale)",
color = "Country"
)
I chose seven large economies from different parts of the world. I used a log scale on the y axis because the values range from under 100 USD to over 90,000 USD; on a normal scale, China and India would look like flat lines at the bottom.
Code
gdp_tidy |>
filter(country %in% selected, year %in% c(1960, 2000, 2015, 2024)) |>
select(country, year, gdp_per_capita) |>
mutate(gdp_per_capita = round(gdp_per_capita)) |>
pivot_wider(names_from = year, values_from = gdp_per_capita, names_prefix = "usd_") |>
arrange(desc(usd_2024))# A tibble: 7 × 5
country usd_1960 usd_2000 usd_2015 usd_2024
<chr> <dbl> <dbl> <dbl> <dbl>
1 United States 3000 36330 56849 86170
2 Germany 1162 23926 41930 56104
3 Japan 519 39753 35665 33797
4 Turkiye 267 4199 11065 15893
5 China 90 969 8175 13293
6 Brazil 235 3767 8936 10311
7 India 85 443 1584 2592This table shows the GDP per capita (in current USD) of the selected countries in four years.
Fastest growth in recent years (2015–2024)
Code
gdp_growth <- gdp_tidy |>
filter(year %in% c(2015, 2024)) |>
select(country, year, gdp_per_capita) |>
pivot_wider(names_from = year, values_from = gdp_per_capita, names_prefix = "usd_") |>
# Keep only countries with data in both years
drop_na() |>
mutate(growth_pct = round((usd_2024 - usd_2015) / usd_2015 * 100, 1))
# Number of countries in the comparison
nrow(gdp_growth)[1] 200Code
# Top 10 fastest-growing countries
top_10 <- gdp_growth |>
slice_max(growth_pct, n = 10)
top_10# A tibble: 10 × 4
country usd_2015 usd_2024 growth_pct
<chr> <dbl> <dbl> <dbl>
1 Guyana 5640. 29675. 426.
2 Moldova 2750. 7579. 176.
3 Albania 4200. 11374. 171.
4 Sao Tome and Principe 1298. 3503. 170.
5 Ukraine 2094. 5392. 157.
6 Armenia 3512. 8556. 144.
7 Bulgaria 7269. 17597. 142.
8 Serbia 5820. 13678. 135
9 Kyrgyz Republic 1106. 2515. 128.
10 Guinea 747. 1695. 127.Code
# 5 countries with the largest decline
gdp_growth |>
slice_min(growth_pct, n = 5)# A tibble: 5 × 4
country usd_2015 usd_2024 growth_pct
<chr> <dbl> <dbl> <dbl>
1 Nigeria 2586. 1084. -58.1
2 Lebanon 7714. 4473. -42
3 Afghanistan 566. 417. -26.3
4 Angola 3642. 2721. -25.3
5 Sudan 1292. 985. -23.8Code
ggplot(top_10, aes(x = reorder(country, growth_pct), y = growth_pct)) +
geom_col(fill = "steelblue") +
coord_flip() +
labs(
title = "Top 10 Countries by GDP per Capita Growth (2015–2024)",
x = "Country",
y = "Growth in GDP per capita (%)"
)
I compared 2015 and 2024 because these are recent years where most countries have data. I did not use 2025, because 14 fewer countries have data for it. I kept only countries with values in both years (200 countries), and calculated the percent change for each one.
Comparing regions and groups
Code
gdp_regions |>
filter(region %in% c("World", "Euro area", "High income", "Low income", "Sub-Saharan Africa"),
!is.na(gdp_per_capita)) |>
ggplot(aes(x = year, y = gdp_per_capita, color = region)) +
geom_line() +
scale_y_log10(labels = scales::comma) +
labs(
title = "GDP per Capita in Selected Regions and Groups (1960-2025)",
x = "Year",
y = "GDP per capita (current USD, log scale)",
color = "Region or group"
)
Code
gdp_regions |>
filter(region %in% c("World", "Euro area", "High income", "Low income", "Sub-Saharan Africa"),
year %in% c(2000, 2024)) |>
select(region, year, gdp_per_capita) |>
mutate(gdp_per_capita = round(gdp_per_capita)) |>
pivot_wider(names_from = year, values_from = gdp_per_capita, names_prefix = "usd_")# A tibble: 5 × 3
region usd_2000 usd_2024
<chr> <dbl> <dbl>
1 Euro area 19563 46424
2 High income 22428 50907
3 Low income 317 809
4 Sub-Saharan Africa 630 1543
5 World 5513 13717I compared the world average with a few regions and income groups. In 2024, GDP per capita was 13,717 USD for the world, 46,424 USD for the Euro area, and 50,907 USD for high-income countries, but only 809 USD for low-income countries. This means the average person in a high-income country produced about 63 times more than in a low-income country.
Conclusion
All seven selected countries had higher GDP per capita in 2024 than in 1960. The United States grew from 3,000 USD to 86,170 USD, and China from 90 USD to 13,293 USD. Japan was the exception in recent years: it dropped from 39,753 USD in 2000 to 33,797 USD in 2024.
From 2015 to 2024, Guyana grew the fastest (426.1%), and Nigeria had the biggest drop (-58.1%). These changes have different causes. Guyana started producing oil at the end of 2019, and its economy grew very quickly after that. Nigeria’s drop does not mean its economy became half as big. In June 2023, Nigeria let its currency, the naira, float freely, and the naira lost a large part of its value against the dollar. Because this data is in current US dollars, a weaker currency makes GDP per capita look much smaller, even if the economy itself did not shrink that much.
The region comparison shows a large gap between rich and poor countries: in 2024, GDP per capita in high-income countries (50,907 USD) was about 63 times higher than in low-income countries (809 USD).
The values are not adjusted for inflation, and many countries have missing data in the early years, so I used 2015-2024 for the growth comparison.
Use of LLM
I used Claude (Anthropic) to help me write and debug the R code for this file and to help draft some of the written explanations. I ran all code myself, and the numbers in this file come from my own output.
References
Anthropic. (2026). Claude (Opus 5.5 version) [Large language model]. https://claude.ai
World Bank. (2026). GDP per capita (current US$) [Data set]. World Development Indicators. https://data.worldbank.org/indicator/NY.GDP.PCAP.CD
U.S. Energy Information Administration. (2024). Country analysis brief: Guyana. https://www.eia.gov/international/content/analysis/countries_long/Guyana/pdf/Guyana%202024%20CAB.pdf
Nairametrics. (2023, June 14). Exchange rate closes at N646. https://nairametrics.com/2023/06/14/exchange-rate-closes-at-n646