Read in 2023 Community Resilience Estimates

Read in the Census Bureau Community Resilience Estimates

Data downloaded from https://www.census.gov/programs-surveys/community-resilience-estimates/data/datasets.html

Many thanks to Katherine Ann Willyard for sharing a copy of her paper “Developing a Data-Driven System for Identifying Vulnerable Neighborhoods” which explained quite clearly why this is a superior vulnerability measure, from a purely statistical viewpoint.

Variable	Description
GEO_ID	A geographic identifier on the type of geography and applicable FIPS codes
STATE	State FIPS code
COUNTY	County FIPS code*
TRACT	Tract FIPS code*
NAME	Geographic Area Name
GEO_LEVEL	Geographic level*
WATER_TRACT	Flag that denotes tracts composed completely of water, where no populations reside*
POPUNI	Total population (excludes adult correctional/juvenile facilities and college dorms)
PRED0_E	Estimated number of individuals with zero components of social vulnerability
PRED0_M	Estimated margin of error
PRED0_PE	Rate of individuals with zero components of social vulnerability
PRED0_PM	Rate margin of error
PRED12_E	Estimated number of individuals with one-two components of social vulnerability
PRED12_M	Estimated margin of error
PRED12_PE	Rate of individuals with one-two components of social vulnerability
PRED12_PM	Rate margin of error
PRED3_E	Estimated number of individuals with three plus components of social vulnerability
PRED3_M	Estimated margin of error
PRED3_PE	Rate of individuals with three plus components of social vulnerability
PRED3_PM	Rate margin of error

How is Vulnerability calculated?

The following ten vulnerability indicators are used in the CRE:

1. Households with an income-to-poverty ratio less than 130 percent
2. Only one or no individuals living in the household are aged 18-64
3. Household crowding defined as more than 0.75 persons per room
4. Household with a communication barrier defined as either limited English-speaking households or households where no one over the age of 16 has a high school diploma
5. No one in the household is employed full-time, year-round, but the flag is not applied if all residents of the household are aged 65 or older
6. Individual with a disability posing a constraint to significant life activity,
7. Individual with no health insurance
8. Individual aged 65 or older
9. Household without a vehicle
10. Households without broadband internet access.

Let’s look at how these numbers are distributed

#   1 or 2 components

df %>% 
  ggplot(aes(x=PRED12_PE)) +
  geom_histogram(bins=200) +
  labs(
    title="Rate of Individuals with 1 or 2 Vulnerability Components",
    subtitle="By Census Tract",
    x="Rate of Social Vulnerability",
    y="Number of Tracts",
    caption="Alan Jackson © 2025. Data from Census Bureau"
  )

#     3 or more components

df %>% 
  ggplot(aes(x=PRED3_PE)) +
  geom_histogram(bins=200) +
  labs(
    title="Rate of Individuals with 3 or More Vulnerability Components",
    subtitle="By Census Tract",
    x="Rate of Social Vulnerability",
    y="Number of Tracts",
    caption="Alan Jackson © 2025. Data from Census Bureau"
  )

#   Number of peolple vs. proportion

df %>% 
  ggplot(aes(x=PRED12_PE, y=PRED12_E)) +
  geom_point() +
  geom_smooth(method = "lm") +
  labs(
    title="Rate of Individuals with 1 or 2 Vulnerability Components vs. Population",
    subtitle="",
    x="Rate of Social Vulnerability",
    y="Number of People in the Category",
    caption="Alan Jackson © 2025. Data from Census Bureau"
  )

df %>% 
  ggplot(aes(x=PRED3_PE, y=PRED3_E)) +
  geom_point() +
  geom_smooth(method = "lm") +
  labs(
    title="Rate of Individuals with 3 or More Vulnerability Components vs. Population",
    subtitle="",
    x="Rate of Social Vulnerability",
    y="Number of People in the Category",
    caption="Alan Jackson © 2025. Data from Census Bureau"
  )

Attach geometry

Attach geometry to the tracts so we can later make a map. Just look at tracts with a P3 ratio > 20, and more than 300 people in P3 in Louisiana.

#   get table of state codes and state abbreviations from tidycensus

data(fips_codes)
# force(fips_codes)

df_map <- NULL

#   Pull out tracts of interest

foo <- df %>% 
  filter(PRED3_E>300 & PRED3_PE>20)  

# Attach state abbreviations to data frame

foo <-
  as_tibble_col(unique(foo$STATE), column_name = "STATE") %>% 
  inner_join(., fips_codes %>% filter(county_code=="001"), 
             join_by(STATE==state_code)) %>% 
  select(STATE, state) %>% 
  left_join(foo, ., by="STATE")

#   Attach geometry

Census <- readRDS(paste0(path, "Census_Tracts_2023/All_ACS_2023.rds"))

foo2 <- foo %>% 
  mutate(GEOID=str_remove(GEO_ID, "^\\d*US")) %>% 
  inner_join(., Census, by="GEOID") %>% 
  sf::st_as_sf()

tmap::tmap_options(basemap.server="OpenStreetMap")
tmap::tmap_mode("view") # set mode to interactive plots

# tmap mode set to interactive viewing
  tmap::tm_shape(foo2 %>% filter(state=="LA")) + 
  tmap::tm_polygons(size=0.5, fill="PRED3_PE", fill_alpha=0.4)

Conclusions

This should be a more robust and more useful vulnerability measure than simply using income or poverty levels. It is a bit of a shame that the numbers for the individual components are not broken out, but I expect that could raise privacy issues. Similarly, it would be nice if it were available at the block-group level, but again, that might raise privacy issues.