choroplethr: Plotting MSA-level maps?

Question

My question relates to plotting a whole U.S. map on MSA level via choroplethr and choroplethrZip.

In the example below, we plot 1) Census population information on a U.S. map on county level and 2) a zoomed map of a selected Metropolitan/Micropolitan Statistical Area (MSA) level.

Example R code:

library(choroplethr)
library(choroplethrZip)

?zip.regions
data(zip.regions)
head(zip.regions)

?df_pop_county
data(df_pop_county)
df_pop_county

?df_pop_zip
data(df_pop_zip)

# U.S. County Population Data
county_choropleth(df_pop_county, legend = "Population")

# NY-NJ-PA MSA Population Data
zip_choropleth(df_pop_zip,
               msa_zoom = "New York-Newark-Jersey City, NY-NJ-PA",
               title    = "2012 NY-Newark-Jersey City MSA\nZCTA Population Estimates",
               legend   = "Population")

Instead of only a zoom into a specific MSA, can we also plot a whole MSA-level U.S. map? An approach like

zip_choropleth(df_pop_zip, legend = "Population")

did not work, and would also likely plot the ZCTA regions, not MSA regions.

Thank you!

Answer 1

You can use the state_zoom argument to zip_choropleth. But as noted in the package documents there is no MSA-based choropleth. An example of how this looks:

states <- unique( zip.regions$state.name) 
lower48 <- states[ ! states %in% c('alaska','hawaii') ]

zip_choropleth(df_pop_zip,
               state_zoom = lower48  ,
               title    = "2012 MSA\nZCTA Population Estimates",
               legend   = "Population")

It looks mostly grey because the ZCTA borders are rendered in grey and they are dense at this scale. If you run the code and look at a higher resolution you can see more of the fill.

An alternative I'd recommend for your task is the tidycensus package. See the code snippet below which I believe creates a map similar to what you're interested in. I select only a few states to clarify the visual, and plot at the county level. I also only plot the top 85 percent of MSAs by total population. This eliminates Danville Virginia for example.

# adapted from https://walkerke.github.io/2017/06/comparing-metros/
library(viridis)
library(ggplot2)
library(tidycensus)
library(tidyverse)
library(tigris)
library(sf)
options(tigris_class = "sf")
options(tigris_use_cache = TRUE)
# census_api_key("YOUR KEY HERE")

acs_var <- 'B01003_001E'
tot <- get_acs(geography = "county", variables = acs_var, state=c("PA", "VA", "DC","MD"),
                 geometry = TRUE)

head(tot)

metros <- core_based_statistical_areas(cb = TRUE) %>%
  select(metro_name = NAME)

wc_tot <- st_join(tot, metros, join = st_within, 
                   left = FALSE) 

pct85 <-  wc_tot %>% group_by(metro_name) %>% 
  summarise(tot_pop=sum(estimate)) %>% summarise(pct85 =  quantile(tot_pop, c(0.85)))
pct85_msas = wc_tot %>% group_by(metro_name) %>% 
  summarise(tot_pop=sum(estimate)) %>% filter(tot_pop > pct85$pct85[1])

head(wc_tot)

ggplot(filter(wc_tot, metro_name %in% pct85_msas$metro_name),
       aes(fill = estimate, color = estimate)) + 
  geom_sf() + 
  coord_sf(crs=3857) + 
  #facet_wrap(~metro_name, scales = "free", nrow = 1) + 
  theme_minimal() + 
  theme(aspect.ratio = 1) + 
  scale_fill_viridis() + 
  scale_color_viridis()

Resulting plot:

The facet line that I've commented out seems to be an area of active devlopment in ggplot. I was getting an error with it, but the source article I mentioned shows how it can be put to good use to display one panel per MSA, which makes a lot of sense. See https://github.com/tidyverse/ggplot2/issues/2651.