Project 2: Population Change in the 10 Most Populous States

Introduction

This dataset was posted by Anastasiia Gmyrina in the Week 5 discussion. It comes from the U.S. Census Bureau’s population estimates and covers the 10 most populous states in 2022, with population counts for 2020, 2021, and 2022. My goal is to see how each state’s population changed over those two years.

Raw Data Structure

library(tidyverse)
library(knitr)

raw <- read_csv(
  "https://raw.githubusercontent.com/0pree/symmetrical-robot/refs/heads/main/census_top10_states_2022_untidy.csv",
  show_col_types = FALSE
)

dim(raw)

[1] 10  5

kable(raw)

Rank	Geographic Area	April 1, 2020 (Estimates Base)	July 1, 2021	July 1, 2022
1	California	39538245	39142991	39029342
2	Texas	29145428	29558864	30029572
3	Florida	21538226	21828069	22244823
4	New York	20201230	19857492	19677151
5	Pennsylvania	13002689	13012059	12972008
6	Illinois	12812545	12686469	12582032
7	Ohio	11799374	11764342	11756058
8	Georgia	10711937	10788029	10912876
9	North Carolina	10439414	10565885	10698973
10	Michigan	10077325	10037504	10034113

The table has 10 rows (one per state) and 5 columns. It is wide because each year is a separate column, and the year is buried inside long headers like “April 1, 2020 (Estimates Base)” and “July 1, 2021”. The column names also have spaces and mixed capitalization. There are no missing values.

One detail worth noting is that the 2020 column is an April estimate, while 2021 and 2022 are July estimates. That means the first interval is about 15 months instead of 12, which slightly affects any year to year comparison.

Tidying the Data

population <- raw %>%
  rename(rank_2022 = Rank, state = `Geographic Area`) %>%
  pivot_longer(
    cols = -c(rank_2022, state),
    names_to = "estimate_date",
    values_to = "population"
  ) %>%
  mutate(year = as.integer(str_extract(estimate_date, "\\d{4}"))) %>%
  select(state, rank_2022, year, estimate_date, population)

kable(head(population, 6))

state	rank_2022	year	estimate_date	population
California	1	2020	April 1, 2020 (Estimates Base)	39538245
California	1	2021	July 1, 2021	39142991
California	1	2022	July 1, 2022	39029342
Texas	2	2020	April 1, 2020 (Estimates Base)	29145428
Texas	2	2021	July 1, 2021	29558864
Texas	2	2022	July 1, 2022	30029572

I renamed the columns to lowercase names without spaces, then reshaped the three date columns into rows so each row is one state and one year. To get a usable year, I pulled the four-digit number out of each header with str_extract(). I kept the original estimate_date text in its own column so the April versus July difference is not lost.

nrow(population) == nrow(raw) * 3

[1] TRUE

The tidy table has 30 rows, which matches 10 states times 3 years, so no rows were lost in the reshape.

Analysis

Change from 2020 to 2022

change <- population %>%
  select(state, year, population) %>%
  pivot_wider(names_from = year, values_from = population, names_prefix = "pop_") %>%
  mutate(
    change = pop_2022 - pop_2020,
    pct_change = round(change / pop_2020 * 100, 2)
  ) %>%
  arrange(desc(pct_change))

kable(
  change,
  format.args = list(big.mark = ","),
  col.names = c("State", "2020", "2021", "2022", "Change", "Percent Change (%)")
)

State	2020	2021	2022	Change	Percent Change (%)
Florida	21,538,226	21,828,069	22,244,823	706,597	3.28
Texas	29,145,428	29,558,864	30,029,572	884,144	3.03
North Carolina	10,439,414	10,565,885	10,698,973	259,559	2.49
Georgia	10,711,937	10,788,029	10,912,876	200,939	1.88
Pennsylvania	13,002,689	13,012,059	12,972,008	-30,681	-0.24
Ohio	11,799,374	11,764,342	11,756,058	-43,316	-0.37
Michigan	10,077,325	10,037,504	10,034,113	-43,212	-0.43
California	39,538,245	39,142,991	39,029,342	-508,903	-1.29
Illinois	12,812,545	12,686,469	12,582,032	-230,513	-1.80
New York	20,201,230	19,857,492	19,677,151	-524,079	-2.59

ggplot(change, aes(x = reorder(state, pct_change), y = pct_change, fill = pct_change > 0)) +
  geom_col() +
  coord_flip() +
  scale_fill_manual(values = c("TRUE" = "steelblue", "FALSE" = "firebrick"), guide = "none") +
  labs(
    title = "Population Change from 2020 to 2022",
    x = NULL, y = "Percent Change (%)"
  ) +
  theme_minimal()

Six of the ten states lost population and four gained. Florida grew the most by percentage (3.28%), followed closely by Texas (3.03%), and Texas added the most people overall, about 884,000. New York lost the most by percentage (2.59%) and also the most people, about 524,000. California lost almost as many people (about 509,000), but because its population is so large, that was only a 1.29% drop.

Growth Pattern by Region

change %>%
  mutate(region = if_else(state %in% c("Texas", "Florida", "Georgia", "North Carolina"), "South", "Other")) %>%
  group_by(region) %>%
  summarize(
    states = n(),
    total_change = sum(change),
    .groups = "drop"
  ) %>%
  kable(format.args = list(big.mark = ","), col.names = c("Region", "States", "Total Change"))

Region	States	Total Change
Other	6	-1,380,704
South	4	2,051,239

All four states that grew are in the South. Together they gained about 2.05 million people, while the other six states lost about 1.38 million. Across all ten states, the total population only grew by about 0.37%, so most of the movement was people shifting between states rather than overall growth.

Year by Year

ggplot(population, aes(x = year, y = population / 1e6, color = state)) +
  geom_line() +
  geom_point() +
  scale_x_continuous(breaks = c(2020, 2021, 2022)) +
  labs(
    title = "Population Estimates by State",
    x = "Year", y = "Population (millions)", color = "State"
  ) +
  theme_minimal()

Because California and Texas are so much larger than the other states, this chart makes the changes in the smaller states hard to see, which is why the percent change chart above is more useful for comparing them.

Conclusions

This was the simplest of the three datasets to tidy. The main step was pulling the year out of long column headers and reshaping the year columns into rows. The analysis shows a clear pattern: between 2020 and 2022, the large Southern states gained population while New York, Illinois, and California lost it. Since the 2020 estimate is from April and the others are from July, the first interval is a little longer, so these numbers slightly overstate change compared to a true two year window.

A next step would be to add more recent Census estimates to see whether this shift continued after 2022, or to include all 50 states and look at domestic migration specifically.

AI Citation

Anthropic. (2026). Claude Opus 5.5 [Large language model]. https://claude.ai. Accessed October 2026.