Reviewing the ACS-1 2023 child poverty estimates

This post looks at how child poverty estimates have changed in the latest U.S. Census data
R
data-viz
tidyverse
tidycensus
api
purrr
Author

Mickey Rafa

Published

September 12, 2024

Introduction

In a previous tidycensus:: post, I showed how to fetch data and do some basic, longitudinal analysis of U.S. Census data. Today, the Census Bureau released the 1-year estimates from the American Community Survey (ACS), so I’d like to see how child poverty seems to be changing from the 2022 to 2023 releases.

Setup

library(tidyverse)
library(tidycensus)
library(ggtext)
library(scales)
library(gt)
library(glue)
library(ggforce)

# census_api_key('INSERT KEY HERE', install = TRUE)
1
Loading the scales:: package to transform ggplot scales simply (some people choose to explicitly define scales:: in their code rather than loading the library).
2
The gt:: library provides functionality for creating ggplot-esque tables.
3
The glue:: package allows for simple addition of HTML to ggplot graphics.
4
The ggforce:: package includes a geom_link() function, which I’ll use to create the comet effect in the comet plots that I use.
5
The first time that you’re working with the tidycensus:: package, you need to request an API key at https://api.census.gov/data/key_signup.html. The install= argument will install your personal key to the .Renviron file, and you won’t need to use the census_api_key() function again.

Data

For this analysis, I’m interested in looking at the most recent state-level child poverty data available from the U.S. Census Bureau, and I want to construct a longitudinal sense of the change in child poverty.

First, let’s revisit the different American Community Survey products – ACS-1 and ACS-5.

What’s the difference between these, and how do you choose which survey product to use for your purposes?

Feature ACS 1-Year Estimates ACS 5-Year Estimates
Data Collection Period 12 months 60 months
Population Coverage Areas with 65,000 or more people All geographic areas, including those with fewer than 65,000 people
Sample Size Smallest Largest
Reliability Less reliable due to smaller sample size More reliable due to larger sample size
Currency Most current data Less current, includes older data
Release Frequency Annually Annually
Best Used For Analyzing large populations, when currency is more important than precision Analyzing small populations, where precision is more important than currency
Example Usage Examining recent economic changes Examining trends in small geographic areas or small population subgroups

For this post, I am interested in constructing a longitudinal dataset of the most recent year-on-year estimates, including the 2023 1-year estimates that were released today. Note that for the ACS-1 products, estimates are only available for geographic units with populations greater than 65,000.

As I did in prior posts, I’ll use the following series from the American Community Survey:

  • B01001_003: Estimate!!Total:!!Male:!!Under 5 years (all racial groups)
  • B01001_027: Estimate!!Total:!!Female:!!Under 5 years (all racial groups)
  • B17001_004: Estimate!!Total:!!Income in the past 12 months below poverty level:!!Male:!!Under 5 years
  • B17001_018: Estimate!!Total:!!Income in the past 12 months below poverty level:!!Female:!!Under 5 years

Fetching from the tidycensus:: API

years <- c(2022, 2023)

Next, I’ll define a function to fetch the ACS-1 data for 2022 and the new 2023 estimates.

fetch_acs_data <- function(year) {
  get_acs(geography = "state", 
          survey = 'acs1',
          variables = c(male_u5_pop = 'B01001_003', 
                        female_u5_pop = 'B01001_027', 
                        male_u5_poverty = 'B17001_004', 
                        female_u5_poverty = 'B17001_018'),
          year = year,
          output = 'wide') %>% 
    mutate(year = year,
           total_u5_popE = male_u5_popE + female_u5_popE,
           total_u5_povertyE = male_u5_povertyE + female_u5_povertyE,
           perc_u5_poverty = total_u5_povertyE / total_u5_popE)
}

fetch_acs_data_county <- function(year) {
  get_acs(geography = "county", 
          survey = 'acs1',
          variables = c(male_u5_pop = 'B01001_003', 
                        female_u5_pop = 'B01001_027', 
                        male_u5_poverty = 'B17001_004', 
                        female_u5_poverty = 'B17001_018'),
          year = year,
          geometry = TRUE,
          output = 'wide') %>% 
    mutate(year = year,
           total_u5_popE = male_u5_popE + female_u5_popE,
           total_u5_povertyE = male_u5_povertyE + female_u5_povertyE,
           perc_u5_poverty = total_u5_povertyE / total_u5_popE,
           county = str_remove(NAME, " County.*"),
           state = str_extract(NAME, "[\\w\\s]+$") %>% str_trim()) %>%
    select(county, state, everything()) %>% 
    group_by(county, state) %>% 
    mutate(perc_u5_poverty_tminus_1 = lag(perc_u5_poverty),
           perc_diff_YoY = perc_u5_poverty - perc_u5_poverty_tminus_1)
}
6
Creating some fields to combine gender-based poverty estimates and calculate a percent of the child population measure
7
Creating state and county fields to manipulate the dataframe and visualize with more simplicity

Then, I’ll use purrr::map_df() to apply each year to the fetch_acs_data() function that I created, which will result in a single dataframe of all years.

combined_acs_data <- map_df(years, fetch_acs_data)
combined_acs_data_county <- map_df(years, fetch_acs_data_county)

Analysis

According to the 2023 ACS-1, roughly 113,000 fewer children were below the Federal Poverty Line.

Estimated child poverty in the U.S. (2022-2023)
Total and Percent estimates of those living below the Federal Poverty Line
Year Total children < 5 y.o. Total children < 5 y.o. living in poverty in last 12 mos. % of children < 5 y.o. living in poverty in last 12 mos.
2022 18,358,199 3,141,107 17.1%
2023 18,333,697 3,027,969 16.5%
Data from 2022 & 2023 American Community Survey 1-year estimates from the U.S. Census Bureau. Estimates exclude Puerto Rico. 
`summarise()` has grouped output by 'year'. You can override using the
`.groups` argument.

Estimated child poverty in the U.S. (2022-2023)
Total and Percent estimates of those living below the Federal Poverty Line
Year Total children < 5 y.o. Total children < 5 y.o. living in poverty in last 12 mos. % of children < 5 y.o. living in poverty in last 12 mos.
Alabama
2022 284,064 71,573 25.2%
2023 288,019 66,642 23.1%
Alaska
2022 46,497 9,199 19.8%
2023 45,211 6,951 15.4%
Arizona
2022 393,413 68,494 17.4%
2023 391,142 62,009 15.9%
Arkansas
2022 177,765 45,650 25.7%
2023 176,908 39,571 22.4%
California
2022 2,118,386 320,754 15.1%
2023 2,086,820 298,927 14.3%
Colorado
2022 305,063 33,128 10.9%
2023 303,775 32,175 10.6%
Connecticut
2022 178,453 23,943 13.4%
2023 180,561 23,506 13.0%
Delaware
2022 54,058 7,072 13.1%
2023 54,398 8,202 15.1%
District of Columbia
2022 39,035 5,729 14.7%
2023 38,512 5,011 13.0%
Florida
2022 1,101,350 210,146 19.1%
2023 1,122,270 183,922 16.4%
Georgia
2022 621,126 110,286 17.8%
2023 621,750 118,763 19.1%
Hawaii
2022 78,927 9,405 11.9%
2023 77,420 8,020 10.4%
Idaho
2022 111,816 15,600 14.0%
2023 110,908 15,449 13.9%
Illinois
2022 674,211 111,480 16.5%
2023 661,026 97,150 14.7%
Indiana
2022 399,031 71,339 17.9%
2023 401,558 67,570 16.8%
Iowa
2022 180,010 24,655 13.7%
2023 182,063 27,260 15.0%
Kansas
2022 176,673 26,764 15.1%
2023 169,830 24,238 14.3%
Kentucky
2022 260,433 58,904 22.6%
2023 264,633 60,196 22.7%
Louisiana
2022 270,937 67,924 25.1%
2023 275,636 72,151 26.2%
Maine
2022 61,018 6,750 11.1%
2023 59,898 6,982 11.7%
Maryland
2022 349,193 39,601 11.3%
2023 346,836 41,543 12.0%
Massachusetts
2022 342,252 36,145 10.6%
2023 342,145 41,153 12.0%
Michigan
2022 536,805 105,043 19.6%
2023 529,459 93,760 17.7%
Minnesota
2022 328,095 37,251 11.4%
2023 326,995 33,516 10.2%
Mississippi
2022 169,303 42,197 24.9%
2023 167,015 37,854 22.7%
Missouri
2022 349,648 64,640 18.5%
2023 348,416 54,655 15.7%
Montana
2022 57,024 7,363 12.9%
2023 55,363 6,976 12.6%
Nebraska
2022 121,107 17,323 14.3%
2023 120,499 14,977 12.4%
Nevada
2022 172,575 29,294 17.0%
2023 171,163 33,530 19.6%
New Hampshire
2022 62,666 5,192 8.3%
2023 62,779 4,565 7.3%
New Jersey
2022 513,333 68,408 13.3%
2023 518,528 64,651 12.5%
New Mexico
2022 104,994 23,304 22.2%
2023 104,293 28,040 26.9%
New York
2022 1,055,455 191,446 18.1%
2023 1,035,708 191,550 18.5%
North Carolina
2022 584,492 105,845 18.1%
2023 594,739 107,284 18.0%
North Dakota
2022 47,844 5,553 11.6%
2023 46,488 3,575 7.7%
Ohio
2022 661,196 123,512 18.7%
2023 654,683 121,143 18.5%
Oklahoma
2022 240,173 50,620 21.1%
2023 239,611 54,783 22.9%
Oregon
2022 199,584 29,443 14.8%
2023 198,150 28,553 14.4%
Pennsylvania
2022 668,734 106,428 15.9%
2023 663,339 111,980 16.9%
Rhode Island
2022 51,955 5,117 9.8%
2023 52,718 7,062 13.4%
South Carolina
2022 281,426 56,398 20.0%
2023 285,830 55,388 19.4%
South Dakota
2022 57,246 10,086 17.6%
2023 54,886 8,240 15.0%
Tennessee
2022 402,215 78,962 19.6%
2023 411,032 80,230 19.5%
Texas
2022 1,881,718 389,244 20.7%
2023 1,913,591 370,590 19.4%
Utah
2022 228,464 23,266 10.2%
2023 229,881 22,545 9.8%
Vermont
2022 27,875 4,312 15.5%
2023 27,168 2,295 8.4%
Virginia
2022 481,682 68,691 14.3%
2023 476,744 59,128 12.4%
Washington
2022 421,722 44,774 10.6%
2023 417,322 55,639 13.3%
West Virginia
2022 87,469 25,753 29.4%
2023 87,453 21,532 24.6%
Wisconsin
2022 309,244 42,078 13.6%
2023 307,874 40,988 13.3%
Wyoming
2022 30,444 5,023 16.5%
2023 30,651 5,549 18.1%
Data from 2022 & 2023 American Community Survey 1-year estimates from the U.S. Census Bureau

To further visualize the state-level changes in estimates, I wanted to try out a comet plot. Comet plots are stylized connected dot plots, which show the “before and after” at both ends of the plot. The comet’s “tail” is the before, and where the estimate moved to is the comet-shaped end. I’ve never used one before, but I think that they’re more than just a fun way to plot data. I think that the movement associated with the comet analogy gives the design some intuition.

I was inspired by this plot, which evaluates NBA player performance using a comet plot. I borrowed heavily from this code and aesthetic.

combined_acs_data %>% 
  filter(year == 2023,
         NAME != 'Puerto Rico') %>% 
  arrange(-perc_diff_YoY) %>% 
  mutate(state = factor(NAME, levels = NAME),
         direction = case_when(
           perc_diff_YoY > .005 ~ "Increase",
           perc_diff_YoY < -.005 ~ "Decrease",
           TRUE ~ "No Change")) %>% 
  ggplot(aes(y = reorder(NAME, -perc_diff_YoY),
             color=direction)) +
  geom_link(aes(x = perc_u5_poverty_tminus_1, y = fct_reorder(NAME, perc_diff_YoY), 
              xend = perc_u5_poverty, yend = fct_reorder(NAME, perc_diff_YoY), 
              size = after_stat(index))) + 
  geom_point(data = . %>% filter(perc_diff_YoY > 0),
             aes(perc_u5_poverty, y = fct_reorder(NAME, perc_diff_YoY)),
             shape = 21, fill = "white", size = 4)  +
  geom_point(data = . %>% filter(perc_diff_YoY < 0),
             aes(perc_u5_poverty, y = fct_reorder(NAME, perc_diff_YoY)),
             shape = 21, fill = "white", size = 4) +
  annotate(geom = 'label', x = .27, y = 49, label = "Poverty rate increased\n in 2023", 
         family = "Roboto Condensed", color = "#E64B35FF", fontface = 'bold', 
         fill = "floralwhite",
         label.size = 0, size = 4) + 
  annotate(geom = 'label', x = .27, y = 36.5, label = "Poverty rate unchanged\n in 2023", 
         family = "Roboto Condensed", color = "#444444", fontface = 'bold', 
         fill = "floralwhite",
         label.size = 0, size = 4) + 
  annotate(geom = 'label', x = .27, y = 24.5, label = "Poverty rate decreased\nin 2023", 
         family = "Roboto Condensed", color = "#00A087FF", fontface = 'bold', 
         fill = "floralwhite",
         label.size = 0, size = 4) + 
  scale_color_manual(values = c("Decrease" = "#00A087FF", "Increase" = "#E64B35FF", "No Change" = "#444444")) +
  scale_size(range = c(.01, 4)) +
  labs(title='Change in State-level Child Poverty<br>from 2022 to 2023',
       subtitle = 'Length of the color band corresponds to the magnitude of change<br>', 
       caption="<br>Source: U.S. Census Bureau's American Community Survey 1-year Estimates<br>ACS-1 has wider undercertainty bands due to smaller sampling",
       x='\n% of children (under 5 y.o.) living\nunder the Federal Poverty Line in 2023\n',
       y='') + 
  scale_x_continuous(labels = percent, position='top') + 
  my_theme + 
  theme(legend.position = 'none',
        plot.title = element_textbox_simple(size=22))

I chose to sort states based on the percentage point change in child poverty, year-over-year. This results in the states with the largest increase in the child poverty rate on top, and states with the largest decrease in the child poverty rate on the bottom. In New Mexico, the child poverty rate increased 4.7 percentage points from the 2022 to the 2023 estimates from the ACS. In Vermont, the poverty rate decreased 7 percentage points from 2022 to 2023. Although West Virginia saw the second largest decrease in child poverty in these estimates, still about 1 out of every four children in the state are below the poverty line, according to the American Community Survey.

It’s important to remember that all ACS estimates have a margin of error (which by default is a 90% confidence interval), so point estimates should be complemented with their respective error bands. ACS-1 is a smaller sampling than the ACS-5 product, so these margins of error can be more significant. This shows that one of New Mexico, Louisiana, or West Virginia could all have the highest under-five poverty rate as of the 2023 survey.

  1. Using the tidycensus:: margin of error aggregation functions to create confidence bands around the point estimate.

A spotlight on Colorado and county-level poverty estimates

Now I want to look into the movement in child poverty in Colorado counties with the ACS-1 2023 estimates. It’s important to remember that ACS-1 estimates are limited to only geographies that exceed 65,000 people. This limits the Colorado data to only 12 counties; however, these 12 counties constitute 88% of the children under five in the whole state.1

Warning: `stat(index)` was deprecated in ggplot2 3.4.0.
ℹ Please use `after_stat(index)` instead.

These changes should also be taken in the context of the margins of error around point estimates. As shown below, the margins of error for some counties’ estimates is quite large. In Mesa County, the margin of error is +/- 10% around the point estimate (of 17.8%).

Conclusion

In this tidycensus:: post, I demonstrated:

  • How to fetch data across multiple years from the U.S. Census Bureau and wrangle the data for longitudinal analysis
  • How to create nice tables using the gt:: package
  • How to construct a comet plot, thanks to some inspiration from a Substack post
  • How to evaluate changes in ACS-1 estimates in the context of their margins of error

More to come on poverty analysis in future posts!

Footnotes

  1. This calculation is done using the 2022 ACS-5 estimates to include all 64 counties in Colorado.↩︎