election_map.md

Here is the process I used to create the 2016 presidential election by county map with zero-population blocks removed.

I obtained US county boundaries and Census block boundaries by state from NHGIS, http://www.nhgis.org, along with Table P1 (Total Population) from the 2010 US Census. Your NHGIS IDs would be different.
I ran the following R code to walk through every block boundary dataset, identify the no-population blocks, then write out the result as a shapefile:

library(tidyverse)
library(sf)
library(tigris)

folders <- list.files("nhgis0079_shape/", pattern = ".zip", 
                      full.names = TRUE)

unzip_and_move <- function(x) {
  unzip(x, exdir = "data/blocks")
}

map(folders, unzip_and_move)

df <- read_csv("nhgis0078_csv/nhgis0078_csv/nhgis0078_ds172_2010_block.csv")

nopop <- df %>%
   select(GISJOIN, totalpop = H7V001) %>%
   filter(totalpop == 0)

state_vec <- unique(fips_codes$state)[1:51]

merge_and_write <- function(state) {  
  message("Processing ", state, "...")  
  layer_name = paste0(state, "_block_2010")  
  blks <- st_read(dsn = "data/blocks", layer = layer_name, 
                  stringsAsFactors = FALSE)  
  blks_nopop <- inner_join(blks, nopop, by = "GISJOIN")
  
  out_name <- paste0(state, "_nopop")
  
  st_write(blks_nopop, dsn = "data/blocks/nopop", layer = out_name, 
           driver = "ESRI Shapefile")  
}

walk(state_vec, merge_and_write)

In ArcGIS Pro, I added all of the no-population block files to a Project and combined them with the Merge tool.
I added the US counties to the Project and used the Erase tool to remove all no-population areas. This takes a very long time (several hours for me) so be patient.
I ran the following R code to obtain 2016 election results, examining Trump votes relative to Clinton votes by county. Note that I'm excluding third-party candidates here and using percentage as percentage of votes cast for either Trump or Clinton.

votes16 <- read_csv("https://raw.githubusercontent.com/mkearney/presidential_election_county_results_2016/master/pres.elect16.results.dec9.csv") %>%
  mutate(fips = paste0("G", 
                       stringr::str_pad(fips, 5, "left", "0"))) %>%
  filter(st != "US" & !is.na(county) & cand %in% c("Donald Trump", "Hillary Clinton")) %>%
  group_by(fips) %>%
  mutate(pct2 = 100 * (votes / sum(votes))) %>%
  ungroup() %>%
  filter(cand == "Donald Trump") 

write_csv(votes16, "pct_trump.csv")

I created a new column in my counties dataset in ArcGIS Pro to match the fips column in the pct_trump.csv dataset, and joined the CSV to the county dataset in ArcGIS Pro.
I created a new layer for Alaska, and gave all boroughs the value of 58.3716, as Alaska only releases statewide results.
I styled the map in ArcMap due to repeated crashes of ArcGIS Pro, using eight categories with breaks at 20, 30, 40, 50, 60, 70, and 80 percent.

walkerke/election_map.md