Part One: Exploring child poverty data with the tidycensus R package

This post uses the tidycensus package to access the Census API and visualize data on child poverty in the U.S.
R
tidyverse
mapping
gt
tidycensus
api
Author

Mickey Rafa

Published

August 13, 2024

Introduction

I’ve used data from the U.S. Census Bureau several times, and for this project, I wanted to reacquaint myself with the tidycensus:: package to gather and wrangle data. I also wanted to use the usmap:: package to generate a simple U.S. map, and the gt:: package to display the data in a nice table format.

Setup

library(tidyverse)
library(tidycensus)
library(scales)
library(janitor)
library(glue)
library(gt)
library(usmap)

# 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 janitor::clean_names() function tidies the column names of your dataset to use the snake case convention. Very handy!
3
The glue:: package allows for simple addition of HTML to ggplot graphics.
4
The gt:: library provides functionality for creating ggplot-esque tables.
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. The tidycensus:: package allows API access to the decennial Census, as well as the more frequent American Community Survey (ACS), which I’ll use in this project.

If you’ve worked with ACS data before, you may know that there are a few survey products offered in the ACS suite. Most commonly, the choice of data is between the 1-year estimates and the 5-year estimates.

What’s the difference between these, and how do you choose which survey product to use for your purposes?1

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


To start, I am interested in reviewing the most stable, geographically-available data on child poverty. Given that I’m less concerned with recency and more interested in broad availability, the ACS 5-year estimates are what I’ll use here.

Initial tour of key tidycensus::get_acs() function

The tidycensus:: package has so much to offer (and I still have plenty to learn!). The tidycensus::load_variables() function provides a simple way to query the available data within each survey. Combining this with stringr::str_detect() is a nice way to search through the tens of thousands of data series that are available through the U.S. Census API.

load_variables(2022, "acs5", cache = TRUE) %>% 
  filter(str_detect(label, "Under 5 years"))
6
This searches all variable labels for “Under 5 years” to help identify data of interest. 
# A tibble: 240 × 4
   name        label                                    concept        geography
   <chr>       <chr>                                    <chr>          <chr>    
 1 B01001A_003 Estimate!!Total:!!Male:!!Under 5 years   Sex by Age (W… tract    
 2 B01001A_018 Estimate!!Total:!!Female:!!Under 5 years Sex by Age (W… tract    
 3 B01001B_003 Estimate!!Total:!!Male:!!Under 5 years   Sex by Age (B… tract    
 4 B01001B_018 Estimate!!Total:!!Female:!!Under 5 years Sex by Age (B… tract    
 5 B01001C_003 Estimate!!Total:!!Male:!!Under 5 years   Sex by Age (A… tract    
 6 B01001C_018 Estimate!!Total:!!Female:!!Under 5 years Sex by Age (A… tract    
 7 B01001D_003 Estimate!!Total:!!Male:!!Under 5 years   Sex by Age (A… tract    
 8 B01001D_018 Estimate!!Total:!!Female:!!Under 5 years Sex by Age (A… tract    
 9 B01001E_003 Estimate!!Total:!!Male:!!Under 5 years   Sex by Age (N… tract    
10 B01001E_018 Estimate!!Total:!!Female:!!Under 5 years Sex by Age (N… tract    
# ℹ 230 more rows

For this demo, I’ll use the following series:

  • 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

There are a bunch of useful helper functions/arguments to assist in fetching data from the Census API. Some noteworthy ones include:

  • Each variable returns the geography, an estimate, and the margin of error (“moe”). Geographies can span from states, regions and the country as a whole, down to areas like school districts, voting districts, census block groups, and many others.
  • survey=: this defines the produce that you’re using of the American Community Survey. Responses can include “acs1”, “acs3”, or (the default) “acs5”.
  • summary_var=: often the variable that you want would be made more meaningful as a ratio or with a demonminator. For example, the number of children in poverty could be useful on its own, but you’re likely to want to see that series as a percent of the total children. With the summary_var argument, you can tell the function which secondary variable you want to grab in the same API call.
  • ouput=wide: related to the above, I wanted to look at child poverty in a way that would require multiple summary variables (e.g. the percent of girls and boys in poverty). Since you can only have one summary variable, output='wide' allows you to grab all of the series that you may need in the same call.
  • geometry=TRUE: this argument returns the geospatial data in tidy format to create quick ggplot-based maps using geom_sf().
df <- get_acs(geography = 'state',
        variables = c(male_u5_pop = 'B01001_003', 
                      female_u5_pop = 'B01001_027', 
                      male_u5_poverty = 'B17001_004', 
                      female_u5_poverty = 'B17001_018'),
        survey = 'acs5',
        year = 2022,
        output = 'wide')
Getting data from the 2018-2022 5-year ACS

Next, I’ll create some fields to combine gender-based poverty estimates and calculate a percent of the child population measure.

df <- df %>% 
  rename(state = NAME) %>% 
  mutate(total_u5_popE = male_u5_popE + female_u5_popE,
         total_u5_povertyE = male_u5_povertyE + male_u5_povertyE,
         perc_u5_in_poverty = total_u5_povertyE / total_u5_popE)

You can combine point estimates for gender-based poverty by simply adding them, but you can’t do the same for margins of error. The estimated margins of error for each estimate are based on their respective samples and the uncertainty in them. To re-weight the margins of error, you take the square root of the sum of squared margins of error:

new_MOE_calc = sqrt(male_u5_povertyM^2 + female_u5_povertyM^2)

The tidycensus::moe_sum() function will calculate the new margin of error for you!

df <- df %>% 
  group_by(state) %>% 
  mutate(total_u5_poverty_MOEcalc = moe_sum(male_u5_povertyM, female_u5_povertyM),
         perc_u5_in_poverty_MOEcalc = total_u5_poverty_MOEcalc / total_u5_popE)

Now, I can visualize the new point estimates and margins of error for child poverty (not specific to gender).

df %>% 
  filter(state != 'Puerto Rico') %>%
  ggplot(
    aes(x=perc_u5_in_poverty, 
        y=reorder(state, perc_u5_in_poverty))) + 
  geom_point(size=3) +
  geom_errorbarh(aes(xmin=perc_u5_in_poverty - perc_u5_in_poverty_MOEcalc, 
                     xmax=perc_u5_in_poverty + perc_u5_in_poverty_MOEcalc),
                 height = 0.4) + 
  labs(y='',
       x='',
       caption=caption_text) + 
  ggtitle(title_text,
          subtitle=subtitle_text) + 
  scale_x_continuous(labels = percent,
                     position = 'top') + 
  my.theme
7
Removing Puerto Rico because it’s not within a U.S. Census region.

Quick mapping with usmap::

The usmap:: package makes rendering a map of the US quick and easy. Though it’s not meant to replace sf:: or packages that allow for more sophisticated maps, it does allow for a quick way to make a U.S. map. For this demo, I’ll plot the state-level poverty data that I collected and manipulated in earlier steps.

This shows that child poverty – those living below the Federal Poverty Line – is concentrated in southern and southeastern states (as a percent of the child population).

(perc_poverty_map <- plot_usmap(regions = 'states',
           data = df,
           values = 'perc_u5_in_poverty') + 
  scale_fill_viridis_c(option = 'inferno', labels = scales::percent_format()) + 
  labs(title = md("**Estimated child poverty in U.S. states in 2022**"),
       subtitle = "as a % of the total child population under 5 y.o.",
       caption = "Source: 2022 American Community Survey") %>%
  theme(legend.position = 'top',
        legend.title = element_blank()))

Great tables with gt::

(df_tbl <- df %>% 
  select(state, region, total_u5_povertyE, total_u5_popE, perc_u5_in_poverty) %>% 
  filter(state != 'Puerto Rico') %>%
  arrange(-perc_u5_in_poverty) %>% 
  # mutate(u5_perc_in_poverty = u5_perc_in_poverty * 100) %>% 
  gt(groupname_col = "region") %>% 
    cols_label(state = 'State',
               total_u5_popE = 'Total children < 5 y.o.',
               total_u5_povertyE = 'Total children < 5 y.o. living in poverty in last 12 mos.',
               perc_u5_in_poverty = '% of children < 5 y.o. living in poverty in last 12 mos.') %>% 
  # formatting numeric fields
  fmt_number(columns = c(total_u5_popE, total_u5_povertyE), decimals = 0, use_seps = TRUE) %>% 
  fmt_percent(columns = perc_u5_in_poverty, decimals = 1) %>%  
  #add table title
  tab_header(title = md("**Estimated child poverty in U.S. states in 2022**"),
             subtitle = "Total and Percent estimates of those living below the Federal Poverty Line") %>% 
  tab_source_note(source_note = "Data from 2022 American Community Survey 5-year estimates from the U.S. Census Bureau") %>% 
  #apply new style to all column headers
  tab_style(
    locations = cells_column_labels(columns = everything()),
    style = list(
      #thick border
      cell_borders(sides = "bottom", weight = px(3)),
      #make text bold
      cell_text(weight = "bold")
    )
  ) %>% 
  #apply different style to title
  tab_style(locations = cells_title(groups = "title"),
            style = list(
              cell_text(weight = "bold", size = 24)
            )) %>% 
  data_color(
    columns = perc_u5_in_poverty,
    palette = viridis::inferno(100)
  ) %>% 
  opt_all_caps() %>% 
  opt_table_font(
    font = list(
      google_font("Chivo"),
      default_fonts()
    )
  ) %>% 
  tab_options(
    #remove border between column headers and title
    column_labels.border.top.width = px(3),
    column_labels.border.top.color = "transparent",
    #remove border around the table
    table.border.top.color = "transparent",
    table.border.bottom.color = "transparent",
    #adjust font sizes and alignment
    source_notes.font.size = 12,
    heading.align = "left"
  )
)
7
Removing Puerto Rico because it’s not within a U.S. Census region.

Estimated child poverty in U.S. states in 2022
Total and Percent estimates of those living below the Federal Poverty Line
State Total children < 5 y.o. living in poverty in last 12 mos. Total children < 5 y.o. % of children < 5 y.o. living in poverty in last 12 mos.
South
Mississippi 53,126 178,246 29.8%
Louisiana 83,556 289,842 28.8%
Alabama 75,754 291,856 26.0%
Arkansas 46,838 181,324 25.8%
West Virginia 21,928 90,380 24.3%
South Carolina 65,876 283,281 23.3%
Kentucky 60,852 266,553 22.8%
Oklahoma 55,918 247,466 22.6%
Tennessee 86,868 402,591 21.6%
Texas 406,500 1,923,422 21.1%
Georgia 132,692 635,299 20.9%
Delaware 10,792 53,990 20.0%
North Carolina 115,116 589,767 19.5%
Florida 210,898 1,104,565 19.1%
District of Columbia 7,204 41,522 17.3%
Virginia 68,948 494,148 14.0%
Maryland 42,630 358,539 11.9%
West
New Mexico 31,806 115,927 27.4%
Arizona 78,418 402,636 19.5%
Nevada 32,000 178,103 18.0%
Montana 10,574 59,003 17.9%
Idaho 18,578 112,576 16.5%
Wyoming 5,240 32,789 16.0%
Alaska 7,684 48,991 15.7%
California 348,758 2,258,308 15.4%
Oregon 31,846 215,756 14.8%
Washington 57,488 440,172 13.1%
Hawaii 10,806 84,552 12.8%
Colorado 39,118 317,189 12.3%
Utah 24,774 239,517 10.3%
Midwest
Ohio 140,956 676,403 20.8%
Michigan 113,090 552,803 20.5%
Indiana 74,890 409,573 18.3%
South Dakota 10,602 58,085 18.3%
Missouri 63,320 360,175 17.6%
Illinois 120,326 721,165 16.7%
Kansas 27,780 180,483 15.4%
Iowa 27,766 189,797 14.6%
Wisconsin 46,366 321,594 14.4%
Nebraska 17,338 127,312 13.6%
North Dakota 7,018 51,717 13.6%
Minnesota 38,206 340,546 11.2%
Northeast
New York 211,592 1,121,872 18.9%
Pennsylvania 117,306 688,571 17.0%
Rhode Island 8,174 54,172 15.1%
New Jersey 76,718 523,995 14.6%
Connecticut 25,786 182,768 14.1%
Maine 8,690 63,182 13.8%
Massachusetts 44,218 351,208 12.6%
Vermont 3,024 28,275 10.7%
New Hampshire 6,306 62,919 10.0%
Data from 2022 American Community Survey 5-year estimates from the U.S. Census Bureau

Deep dive: Colorado results

df_colorado <- get_acs(geography = 'county',
                       # using the state argument to only fetch Colorado
                       state = "Colorado",
                       survey = 'acs5',
                       variables = c(male_u5_pop = 'B01001_003', 
                                     female_u5_pop = 'B01001_027', 
                                     male_u5_poverty = 'B17001_004', 
                                     female_u5_poverty = 'B17001_018'),
                       year = 2022,
                       output = 'wide')
Getting data from the 2018-2022 5-year ACS

(df_tbl2 <- df_colorado %>% 
  select(county, region, total_u5_povertyE, total_u5_popE, perc_u5_in_poverty) %>% 
  arrange(-perc_u5_in_poverty) %>% 
  gt(groupname_col = "region") %>% 
  cols_label(county = 'County',
               total_u5_popE = 'Total children < 5 y.o.',
               total_u5_povertyE = 'Total children < 5 y.o. living in poverty in last 12 mos.',
               perc_u5_in_poverty = '% of children < 5 y.o. living in poverty in last 12 mos.') %>% 
  # formatting numeric fields
  fmt_number(columns = c(total_u5_popE, total_u5_povertyE), decimals = 0, use_seps = TRUE) %>% 
  fmt_percent(columns = perc_u5_in_poverty, decimals = 1) %>%  
  #add table title
  tab_header(title = md("**Estimated child poverty in Colorado by county in 2022**"),
             subtitle = "Total and Percent estimates of those living below the Federal Poverty Line") %>% 
  tab_source_note(source_note = "Data from 2022 American Community Survey 5-year estimates from the U.S. Census Bureau") %>% 
  #apply new style to all column headers
  tab_style(
    locations = cells_column_labels(columns = everything()),
    style = list(
      #thick border
      cell_borders(sides = "bottom", weight = px(3)),
      #make text bold
      cell_text(weight = "bold")
    )
  ) %>% 
  #apply different style to title
  tab_style(locations = cells_title(groups = "title"),
            style = list(
              cell_text(weight = "bold", size = 24)
            )) %>% 
  data_color(
    columns = perc_u5_in_poverty,
    palette = viridis::inferno(100)
  ) %>% 
  opt_all_caps() %>% 
  opt_table_font(
    font = list(
      google_font("Chivo"),
      default_fonts()
    )
  ) %>% 
  tab_options(
    #remove border between column headers and title
    column_labels.border.top.width = px(3),
    column_labels.border.top.color = "transparent",
    #remove border around the table
    table.border.top.color = "transparent",
    table.border.bottom.color = "transparent",
    #adjust font sizes and alignment
    source_notes.font.size = 12,
    heading.align = "left"
  )
)

Estimated child poverty in Colorado by county in 2022
Total and Percent estimates of those living below the Federal Poverty Line
County Total children < 5 y.o. living in poverty in last 12 mos. Total children < 5 y.o. % of children < 5 y.o. living in poverty in last 12 mos.
Southern
Otero 782 1,053 74.3%
Bent 160 227 70.5%
Huerfano 184 325 56.6%
Kiowa 50 123 40.7%
Pueblo 2,580 9,249 27.9%
Las Animas 178 666 26.7%
Prowers 204 792 25.8%
Conejos 108 471 22.9%
Baca 38 168 22.6%
Crowley 36 173 20.8%
Saguache 62 306 20.3%
Rio Grande 118 589 20.0%
Costilla 24 166 14.5%
Alamosa 80 967 8.3%
Mineral 0 31 0.0%
Mountain
Chaffee 274 767 35.7%
Fremont 488 2,014 24.2%
Clear Creek 46 271 17.0%
Grand 76 652 11.7%
Eagle 284 2,744 10.3%
Teller 92 938 9.8%
Park 28 598 4.7%
Summit 14 1,026 1.4%
Custer 0 80 0.0%
Gilpin 0 202 0.0%
Jackson 0 47 0.0%
Lake 0 313 0.0%
Pitkin 0 807 0.0%
Western
Montezuma 398 1,229 32.4%
Delta 370 1,376 26.9%
La Plata 444 2,484 17.9%
Moffat 134 805 16.6%
Mesa 1,208 8,212 14.7%
Garfield 374 3,975 9.4%
Montrose 188 2,067 9.1%
Archuleta 44 567 7.8%
Routt 44 1,028 4.3%
San Miguel 8 229 3.5%
Rio Blanco 10 415 2.4%
Dolores 0 64 0.0%
Gunnison 0 577 0.0%
Hinsdale 0 23 0.0%
Ouray 0 145 0.0%
San Juan 0 16 0.0%
Eastern
Washington 58 217 26.7%
Morgan 476 2,065 23.1%
Yuma 112 684 16.4%
Logan 182 1,168 15.6%
Kit Carson 68 465 14.6%
Lincoln 30 253 11.9%
Sedgwick 14 190 7.4%
Elbert 64 1,125 5.7%
Phillips 16 336 4.8%
Cheyenne 0 118 0.0%
Front Range
Adams 4,928 34,100 14.5%
Denver 5,386 39,520 13.6%
El Paso 6,040 46,014 13.1%
Arapahoe 4,618 38,209 12.1%
Weld 2,606 22,760 11.4%
Boulder 1,464 13,358 11.0%
Larimer 1,446 16,353 8.8%
Jefferson 1,992 28,086 7.1%
Broomfield 86 3,509 2.5%
Douglas 404 19,682 2.1%
Data from 2022 American Community Survey 5-year estimates from the U.S. Census Bureau

Conclusion

In this first tidycensus:: post, I demonstrated:

  • How to fetch data from the U.S. Census Bureau
  • A simple way to search for the type of data that you’re interested in exploring
  • How to use some of the tidycensus:: functions and arguments to support in data wrangling, including re-weighting the margin of error when combining estimates
  • How to make a simple map using the usmap:: package
  • How to make a clean and visually appealing table using the gt:: package

Footnotes

  1. See the Census handbook for great detail: https://www.census.gov/programs-surveys/acs/library/handbooks/general.html.↩︎