Show R code
library(tidyverse)
library(knitr)
# Each report must run from a fresh R session.
dir.create("data/tidy", recursive = TRUE, showWarnings = FALSE)Project 2 Census Educational Attainment (2024)
Approach
I used the 2024 Census education table (B15003). The file I downloaded has numbers for the whole United States, so I compared education levels nationally. I also kept the margins of error in the data. I did not compare states because this file does not have state-level numbers.
Load packages and create output folders
Source and data before tidying
Official table: 2024 ACS B15003. I kept the original CSV under data/raw/. It contains the category labels and two wide measurement columns: the estimate and its margin of error.
Show R code
# Read the original Census CSV as text first so commas and margin-of-error symbols stay intact.
source_file <- "data/raw/01_census_acs_2024_education.csv"
stopifnot(file.exists(source_file))
raw <- read_csv(source_file, col_types = cols(.default = col_character()))
# Check the exact columns before doing any cleaning. This avoids using the wrong export.
expected <- c("Label (Grouping)", "United States!!Estimate", "United States!!Margin of Error")
stopifnot(identical(names(raw), expected), nrow(raw) == 25L)
cat("Original rows:", nrow(raw), " Columns:", ncol(raw), "\n")Original rows: 25 Columns: 3 Show R code
kable(head(raw, 6), caption = "Original Census data (unchanged)")| Label (Grouping) | United States!!Estimate | United States!!Margin of Error |
|---|---|---|
| Total: | 230,807,303 | ±9,706 |
| No schooling completed | 4,534,710 | ±27,246 |
| Nursery school | 65,861 | ±2,200 |
| Kindergarten | 64,338 | ±2,332 |
| 1st grade | 106,237 | ±3,702 |
| 2nd grade | 220,340 | ±5,639 |
Transform the wide data
The original file has an estimate column and a margin-of-error column. I used pivot_longer() to put them into one value column, with another column showing which measurement it is. I cleaned the numbers after that. I kept the two types separate because the margin of error is not a population count.
Show R code
# Start from the raw wide table; keep estimates separate from their margins of error.
tidy <- raw %>%
# Use a clear variable name for the education categories.
rename(education = `Label (Grouping)`) %>%
# Remove extra/nonbreaking spaces from the downloaded labels.
mutate(education = str_squish(str_replace_all(education, "\u00a0", " "))) %>%
# Stack the two measurement columns instead of treating them as different people.
pivot_longer(
cols = c(`United States!!Estimate`, `United States!!Margin of Error`),
names_to = "measure", values_to = "reported_value"
) %>%
# Give the measure readable labels and turn formatted numbers into numeric values.
mutate(
measure = recode(measure,
`United States!!Estimate` = "estimate",
`United States!!Margin of Error` = "margin_of_error"),
value = readr::parse_number(reported_value)
) %>%
select(education, measure, value)
# Each original education row must now have one estimate and one margin-of-error row.
stopifnot(nrow(tidy) == nrow(raw) * 2, !anyNA(tidy$value),
n_distinct(tidy$education) == nrow(raw))
dir.create("data/tidy", recursive = TRUE, showWarnings = FALSE)
# Save a reusable tidy CSV; the setup chunk creates this folder when needed.
write_csv(tidy, "data/tidy/01_census_education_tidy.csv")
kable(head(tidy, 8), caption = "Tidy data: one educational category and one measurement per row")| education | measure | value |
|---|---|---|
| Total: | estimate | 230807303 |
| Total: | margin_of_error | 9706 |
| No schooling completed | estimate | 4534710 |
| No schooling completed | margin_of_error | 27246 |
| Nursery school | estimate | 65861 |
| Nursery school | margin_of_error | 2200 |
| Kindergarten | estimate | 64338 |
| Kindergarten | margin_of_error | 2332 |
Analyze the tidy data
For the percentages, I used only the rows marked estimate. I took the total number of adults from the Total: row. The percentages show the highest education level reported for adults age 25 and older. I did not add the margins of error to the counts.
Show R code
# Use the published total adult population as the denominator for every percentage.
total_adults <- tidy %>%
filter(education == "Total:", measure == "estimate") %>%
pull(value)
stopifnot(length(total_adults) == 1L, total_adults > 0)
# Analyze estimates only; margins of error are not population counts.
shares <- tidy %>%
filter(measure == "estimate", education != "Total:") %>%
mutate(percent_adults = 100 * value / total_adults)
# The detailed attainment categories should sum to the reported national total.
stopifnot(abs(sum(shares$value) - total_adults) < 0.1)
selected <- c("Regular high school diploma", "Associate's degree",
"Bachelor's degree", "Master's degree", "Doctorate degree")
comparison <- shares %>%
filter(education %in% selected) %>%
arrange(desc(percent_adults))
stopifnot(nrow(comparison) == 5L)
kable(comparison, digits = 2,
col.names = c("Highest education completed", "Measurement", "Adults (people)", "Adults (%)"))| Highest education completed | Measurement | Adults (people) | Adults (%) |
|---|---|---|---|
| Regular high school diploma | estimate | 51011337 | 22.10 |
| Bachelor’s degree | estimate | 49868171 | 21.61 |
| Master’s degree | estimate | 23264421 | 10.08 |
| Associate’s degree | estimate | 20322913 | 8.81 |
| Doctorate degree | estimate | 3875973 | 1.68 |
Show R code
# Based on the R Graph Gallery horizontal bar example: flip long labels, color each category,
# and print the actual Census percentages at the end of the bars.
ggplot(comparison, aes(x = reorder(education, percent_adults), y = percent_adults, fill = education)) +
geom_col(width = .7, color = "white", linewidth = 0.35) +
geom_text(aes(label = sprintf("%.1f%%", percent_adults)),
hjust = -0.15, size = 3.5, show.legend = FALSE) +
scale_y_continuous(expand = expansion(mult = c(0, 0.16))) +
scale_fill_manual(values = c(
"Regular high school diploma" = "#2671AF",
"Associate's degree" = "#137E8A",
"Bachelor's degree" = "#E29B32",
"Master's degree" = "#7A5195",
"Doctorate degree" = "#C44E52"
)) +
coord_flip() +
guides(fill = "none") +
labs(title = "Selected educational attainment levels in the United States",
subtitle = "2024 ACS five-year estimates, adults age 25 and older",
x = NULL, y = "Share of adults age 25+ (%)",
caption = "Source: U.S. Census Bureau, ACS B15003; selected categories only") +
theme_minimal()
Conclusion
Show R code
get_share <- function(x) comparison$percent_adults[match(x, comparison$education)]
cat(sprintf("The Census file reports **%s adults age 25 or older** in the U.S. A regular high school diploma is the highest level for **%.2f%%**, a bachelor's degree for **%.2f%%**, and a master's degree for **%.2f%%**. I divided each count by the full adult total, not just the five categories I selected. I kept the margins of error but did not count them as people. These five categories are selected examples, not a breakdown of all education levels.\n",
format(total_adults, big.mark=",", scientific=FALSE),
get_share("Regular high school diploma"),
get_share("Bachelor's degree"),
get_share("Master's degree")))The Census file reports 230,807,303 adults age 25 or older in the U.S. A regular high school diploma is the highest level for 22.10%, a bachelor’s degree for 21.61%, and a master’s degree for 10.08%. I divided each count by the full adult total, not just the five categories I selected. I kept the margins of error but did not count them as people. These five categories are selected examples, not a breakdown of all education levels.
Data checks and limits
The 24 education categories add up to the published total for adults age 25 and older. I checked that before computing the percentages. The graph shows only five selected categories, so its bars are not supposed to add up to 100%. The ACS margins of error were kept in the tidy file but are not the same as additional adults; this chart shows point estimates, not confidence intervals.
Visualization reference
The horizontal-bar layout and custom category colors are adapted from the R Graph Gallery’s ggplot2 bar-chart examples. The DataDaft video is an additional tutorial reference for bar charts and coord_flip(). The graph is calculated from the Census CSV, not the tutorial data.
Reference
U.S. Census Bureau. (2024). 2024 American Community Survey five-year estimates: Educational attainment (B15003). https://data.census.gov/table/ACSDT5Y2024.B15003
The R Graph Gallery. (n.d.). Basic barplot with ggplot2. https://r-graph-gallery.com/218-basic-barplots-with-ggplot2
DataDaft. (2019, October 14). How to make a bar plot in R [Video]. YouTube. https://www.youtube.com/watch?v=XBqnL2RUVcg
AI assistance
I used ChatGPT to help organize and review the R and Quarto code and to draft presentation notes. The figures come from the original Census CSV. I checked the code decisions against the source fields.