2020 Election Maps

ggplot2
tutorial
maps
Creating 2020 election maps using ggplot2
Author

Andrew Weatherman

Published

June 10, 2024

The What

This tutorial walks through how to create various election maps using ggplot2 and open-source election data from MIT and the North Carolina State Board of Elections. Admittedly, much of the plotting process will be the same. Once you understand how to make this kind of map, it’s fairly easy to extend the workflow to other data.

There are lots of cool maps to create in R, and over the coming weeks, I intended to post a few more articles outlining different options.

What we will be creating NC Gov. Race

Presidential, County Race

Presidential, NC Race

The How

For this table, we will need:

For starters, we’re going to define a base theme that every plot will use.

theme_politics <- function() {
  ggthemes::theme_map() +
    theme(
      plot.title.position = "plot",
      plot.title = element_text(family = "RadioCanadaBig-Bold", hjust = 0.5, size = 24, vjust = 0),
      plot.subtitle = element_text(
        family = "RadioCanadaBig-Regular", hjust = 0.5,
        vjust = -1, size = 16
      ),
      plot.caption.position = "plot",
      plot.caption = element_text(family = "RadioCanadaBig-Regular", size = 12, hjust = 0.5),
      legend.position = "none",
      plot.margin = margin(
        t = 20,
        r = 20,
        b = 20,
        l = 20, unit = "pt"
      ),
      plot.background = element_rect(fill = "#F6F7F2")
    )
}

Gov. Race

Plotting the 2020 North Carolina Governor race.

The Data

To speed things up, I took the full data set from NCSBE, converted it to .parquet, and uploaded it to HuggingFace for retrieval. By default, NCSBE returns data in .txt (hello, it’s 2024 ???) which can take a while to download and index. This will be much quicker.

gov_data <- cbbdata:::parquet_from_url("https://huggingface.co/datasets/andreweatherman/election-maps/resolve/main/results_pct_20201103.parquet?download=true")
County Election Date Precinct Contest Group ID Contest Type Contest Name Choice Choice Party Vote For Election Day One Stop Absentee by Mail Provisional Total Votes Real Precinct …16
TYRRELL 11/03/2020 3 1359 S NC TREASURER Ronnie Chatterji DEM 1 16 0 0 0 16 Y NA
WAYNE 11/03/2020 06 1011 S NC COURT OF APPEALS JUDGE SEAT 05 Fred Gore REP 1 95 0 0 0 95 Y NA
FORSYTH 11/03/2020 081 1342 S NC SUPERINTENDENT OF PUBLIC INSTRUCTION Jen Mangrum DEM 1 204 1076 367 7 1654 Y NA
NEW HANOVER 11/03/2020 W03 1373 S US PRESIDENT Donald J. Trump REP 1 71 0 31 1 103 Y NA
MECKLENBURG 11/03/2020 241 1008 S NC COMMISSIONER OF INSURANCE Mike Causey REP 1 348 2257 0 0 2605 Y NA
TRANSYLVANIA 11/03/2020 PF 1009 S NC COMMISSIONER OF LABOR Jessica Holmes DEM 1 28 225 102 0 355 Y NA

Cleaning

To “clean” our data, we want to do a few things.

  • Standardize column names with janitor::clean_names for easier typing.
  • Filter down to just the Governor race.
  • Sum across all votes in each county to retrieve the total number of county votes.
  • Count the total votes for each candidate, in each county, and compute the percentage won.
  • Filter down to Roy Cooper.

For plotting purposes, our county names need to be lowercase. We also will be plotting relative to a single candidate’s proportion, which will make more sense later, so we only need to chose between one major candidate. Your choice is largely irrelevant and will only affect the direction of your color scale when plotting. Because voters can cast ballots for third-party candidates, we wait to filter until the last step.

gov_data <- gov_data %>%
  janitor::clean_names() %>% 
  filter(contest_name == "NC GOVERNOR") %>% 
    mutate(total_county_votes = sum(total_votes),
         county = tolower(county),
         .by = county) %>% 
  summarize(total = sum(total_votes), 
            prop = total / first(total_county_votes),
            .by = c(choice, county)) %>% 
  filter(choice == "Roy Cooper")
choice county total prop
Roy Cooper durham 147110 0.8191208
Roy Cooper orange 65042 0.7668691
Roy Cooper hertford 7212 0.6824375
Roy Cooper mecklenburg 382726 0.6780716
Roy Cooper edgecombe 16786 0.6595937
Roy Cooper warren 6716 0.6523555

The plot

To plot U.S. maps in ggplot2, I recommend taking advantage of map_data and geom_polygon. map_data returns data from the maps package as a data frame suitable for plotting. You’ll need to join this output with your data. For this example, we only care about North Carolina, so we filter our region to “north carolina.” Importantly, we need to use join_by because our joining “keys” are mismatched (“county” vs. “subregion”).

County Lines

Secondly, to draw county lines we can use geom_path and, again, pull county data from map_data. We can also set border colors (color) and border thickness (linewidth).

Colors

As mentioned above, the candidate you filter on doesn’t matter…but you might need to adjust your color scale depending on their political party. In our example, we chose Roy Cooper, a Democrat, and we are using the “Red-Blue Diverging” palette from ggthemes. If your chosen candidate is Republican, you will need to add direction = -1 inside of scale_fill_paletteer_c to “flip” the scale and have red, instead, as the high point.

Aside from those points, this plot is pretty straightforward. Make sure you have the Radio Canada Big font installed on your machine. You can grab it for free from Google Fonts.

gov_data_plot <- gov_data %>% 
  left_join(map_data("county", region = 'north carolina'), join_by("county" == "subregion")) %>% 
  ggplot(aes(long, lat, fill = prop, group = group)) +
  geom_polygon() +
  geom_path(data = map_data("county", region = 'north carolina'), aes(long, lat, group = group), 
            color = "black", linewidth = 0.4, inherit.aes = FALSE) +
  paletteer::scale_fill_paletteer_c("ggthemes::Red-Blue Diverging") +
  theme_politics() +
  labs(title = "2020 Governor Election Results, North Carolina",
       subtitle = "Roy Cooper (D) won North Carolina by 248,185 votes (51.5% to 47%)",
       caption = "Data by North Carolina State Board of Elections || Viz. by @andreweatherman")

Saving

ggsave(plot = gov_data_plot, "viz/election-plots-2020/nc-gov.png", device = grDevices::png, w = 12, dpi = 600)

National Race, County

Plotting the 2020 Presidential Election on a per-county, national level.

The Data

Like the Governor data, I pulled this from MIT and also uploaded it to HuggingFace. One note: I think this data is missing Oglala Lakota county in South Dakota. If you’re producing these maps in 2024, please ensure that your data is both accurate and complete.

county_data <- cbbdata:::parquet_from_url("https://huggingface.co/datasets/andreweatherman/election-maps/resolve/main/election_2020_county.parquet?download=true") %>% 
  filter(year == 2020)
year state state_po county_name county_fips office candidate party candidatevotes totalvotes version mode
2020 ALABAMA AL AUTAUGA 1001 US PRESIDENT JOSEPH R BIDEN JR DEMOCRAT 7503 27770 20220315 TOTAL
2020 ALABAMA AL AUTAUGA 1001 US PRESIDENT OTHER OTHER 429 27770 20220315 TOTAL
2020 ALABAMA AL AUTAUGA 1001 US PRESIDENT DONALD J TRUMP REPUBLICAN 19838 27770 20220315 TOTAL
2020 ALABAMA AL BALDWIN 1003 US PRESIDENT JOSEPH R BIDEN JR DEMOCRAT 24578 109679 20220315 TOTAL
2020 ALABAMA AL BALDWIN 1003 US PRESIDENT OTHER OTHER 1557 109679 20220315 TOTAL
2020 ALABAMA AL BALDWIN 1003 US PRESIDENT DONALD J TRUMP REPUBLICAN 83544 109679 20220315 TOTAL

Cleaning

To “clean” our data, we want to do a few things.

  • Calculate the proportion of total votes won in a county – summing candidatevotes and dividing by totalvotes – for each candidate in each county.
  • Like above, filter to a single candidate.
  • Bring on county geographic information by joining on fips codes.
county_data <- county_data %>% 
  reframe(prop = sum(candidatevotes) / totalvotes,
          .by = c(candidate, county_fips)) %>% 
  filter(candidate == "JOSEPH R BIDEN JR") %>% 
  left_join(read_rds("https://huggingface.co/datasets/andreweatherman/election-maps/resolve/main/county_map_fips.rds?download=true"), join_by("county_fips" == "fips"))
candidate county_fips prop long lat group order region subregion
JOSEPH R BIDEN JR 1001 0.2701837 -86.50517 32.34920 1 1 alabama autauga
JOSEPH R BIDEN JR 1001 0.2701837 -86.53382 32.35493 1 2 alabama autauga
JOSEPH R BIDEN JR 1001 0.2701837 -86.54527 32.36639 1 3 alabama autauga
JOSEPH R BIDEN JR 1001 0.2701837 -86.55673 32.37785 1 4 alabama autauga
JOSEPH R BIDEN JR 1001 0.2701837 -86.57966 32.38357 1 5 alabama autauga
JOSEPH R BIDEN JR 1001 0.2701837 -86.59111 32.37785 1 6 alabama autauga

The plot

Like mentioned at the start of this post, most of the plotting workflow is the same, so I’ll spare repeating the same information as above.

State Lines

To draw state borders, we’re going to use the same process as before but switch to pulling map_data("state") while also thickening the borders.

county_plot <- county_data %>% 
  ggplot(aes(long, lat, fill = prop, group = group)) +
  geom_polygon() +
  geom_path(data = map_data("state"), aes(long, lat, group = group), 
            color = "black", linewidth = 0.4, inherit.aes = FALSE) +
  geom_path(data = map_data("county"), aes(long, lat, group = group), 
            color = "black", linewidth = 0.1, inherit.aes = FALSE) +
  paletteer::scale_fill_paletteer_c("ggthemes::Red-Blue Diverging") +
  ggthemes::theme_map() +
  theme(plot.title.position = "plot",
        plot.title = element_text(family = "RadioCanadaBig-Bold", hjust = 0.5, size = 24, vjust = 0),
        plot.subtitle = element_text(family = "RadioCanadaBig-Regular", hjust = 0.5,
                                     vjust = -1, size = 16),
        plot.caption.position = "plot",
        plot.caption = element_text(family = "RadioCanadaBig-Regular", size = 14, hjust = 0.5),
        legend.position = "none",
        plot.margin = margin(
          t = 20,
          r = 20,
          b = 20,
          l = 20, unit = "pt"),
        plot.background = element_rect(fill = "#F6F7F2")) +
  labs(title = "2020 Presidential Election Results, Joe Biden Wins (306-232)",
       subtitle = "Joe Biden wins by 7.1M popular votes (51.3% to 46.9%) and 306-232 in the Electoral College",
       caption = "Data by MIT Election Lab || Viz. by @andreweatherman")

Saving

ggsave(plot = county_plot, "viz/election-plots-2020/county.png", device = grDevices::png, w = 13, h = 8.5, dpi = 600)

National Race, State

Plotting the 2020 Presidential Election on a per-county level in North Carolina.

The Data

state_data <- cbbdata:::parquet_from_url("https://huggingface.co/datasets/andreweatherman/election-maps/resolve/main/election_2020_county.parquet?download=true") %>% 
  filter(state_po == "NC" & year == 2020)
year state state_po county_name county_fips office candidate party candidatevotes totalvotes version mode
2020 NORTH CAROLINA NC ALAMANCE 37001 US PRESIDENT JOSEPH R BIDEN JR DEMOCRAT 10159 86091 20220315 ABSENTEE BY MAIL
2020 NORTH CAROLINA NC ALAMANCE 37001 US PRESIDENT JOSEPH R BIDEN JR DEMOCRAT 5264 86091 20220315 ELECTION DAY
2020 NORTH CAROLINA NC ALAMANCE 37001 US PRESIDENT JOSEPH R BIDEN JR DEMOCRAT 23256 86091 20220315 ONE STOP
2020 NORTH CAROLINA NC ALAMANCE 37001 US PRESIDENT JOSEPH R BIDEN JR DEMOCRAT 146 86091 20220315 PROVISIONAL
2020 NORTH CAROLINA NC ALAMANCE 37001 US PRESIDENT OTHER GREEN 50 86091 20220315 ABSENTEE BY MAIL
2020 NORTH CAROLINA NC ALAMANCE 37001 US PRESIDENT OTHER GREEN 55 86091 20220315 ELECTION DAY

Cleaning

Largely the same process as the previous plot.

state_data <- state_data %>% 
  reframe(total_votes = sum(candidatevotes),
          prop = total_votes / totalvotes,
          .by = c(candidate, county_fips)) %>% 
  filter(candidate == "JOSEPH R BIDEN JR") %>% 
  left_join(read_rds("https://huggingface.co/datasets/andreweatherman/election-maps/resolve/main/county_map_fips.rds?download=true"), join_by("county_fips" == "fips"))
candidate county_fips total_votes prop long lat group order region subregion
JOSEPH R BIDEN JR 37001 38825 0.4509763 -79.53800 35.84424 1857 54915 north carolina alamance
JOSEPH R BIDEN JR 37001 38825 0.4509763 -79.54372 35.89008 1857 54916 north carolina alamance
JOSEPH R BIDEN JR 37001 38825 0.4509763 -79.53800 35.98175 1857 54917 north carolina alamance
JOSEPH R BIDEN JR 37001 38825 0.4509763 -79.52081 36.23385 1857 54918 north carolina alamance
JOSEPH R BIDEN JR 37001 38825 0.4509763 -79.26298 36.23385 1857 54919 north carolina alamance
JOSEPH R BIDEN JR 37001 38825 0.4509763 -79.27444 35.90726 1857 54920 north carolina alamance

The plot

Just as the previous plot.

nc_pres <- state_data %>% 
  ggplot(aes(long, lat, fill = prop, group = group)) +
  geom_polygon() +
  geom_path(data = map_data("county", region = 'north carolina'), aes(long, lat, group = group), 
            color = "black", linewidth = 0.4, inherit.aes = FALSE) +
  paletteer::scale_fill_paletteer_c("ggthemes::Red-Blue Diverging") +
  theme_politics() +
  labs(title = "2020 Presidential Election Results, North Carolina",
       subtitle = "Donald Trump won North Carolina and its 15 electoral votes by 74,483 votes (49.9% to 48.6%)",
       caption = "Data by MIT Election Lab || Viz. by @andreweatherman")

Saving

ggsave(plot = nc_pres, "viz/election-plots-2020/nc-pres-state.png", device = grDevices::png, w = 12, dpi = 600)