US Presidential Elections 2020

Mapping Elections Data I

Gaston Sanchez

Required Packages

The content in these slides depend on the following packages:

library(tidyverse)      # data wrangling and graphics

library(sf)             # for working with geospatial vector-data

library(rnaturalearth)  # maps data (e.g. US States)

library(maps)           # maps data (e.g. US Counties)

library(plotly)         # interactive plots

About

Visualizing US Presidential Elections (2020)

US Presidential Elections 2020

Data from MIT Election Lab

http://electionlab.mit.edu/

County Presidential Election (2000-2020)

https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/VOQCHQ

Citation: Data Source

Data: County Presidential Election Returns 2000-2020

Source: MIT Election Data + Science Lab

https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/VOQCHQ

MIT Election Data and Science Lab, 2018, “County Presidential Election Returns 2000-2020”, https://doi.org/10.7910/DVN/VOQCHQ, Harvard Dataverse, V11, UNF:6:HaZ8GWG8D2abLleXN3uEig== [fileUNF]

License: Public Domain CC0 1.0

Data Available in CSV format

US Presidential Election 2000-2020 Data

CSV file available in bCourses (see Files/data)

dat = read_csv(
  file = "countypres_2000-2020.csv", 
  col_types = c("icccccccdddc")) |>
  mutate(
    state = tolower(state),
    county_name = tolower(county_name)
  )

US Presidential Election 2000-2020 Data

# first 6 columns
dat |> select(1:6) |> slice_head(n = 8)
# A tibble: 8 × 6
   year state   state_po county_name county_fips office      
  <int> <chr>   <chr>    <chr>       <chr>       <chr>       
1  2000 alabama AL       autauga     01001       US PRESIDENT
2  2000 alabama AL       autauga     01001       US PRESIDENT
3  2000 alabama AL       autauga     01001       US PRESIDENT
4  2000 alabama AL       autauga     01001       US PRESIDENT
5  2000 alabama AL       baldwin     01003       US PRESIDENT
6  2000 alabama AL       baldwin     01003       US PRESIDENT
7  2000 alabama AL       baldwin     01003       US PRESIDENT
8  2000 alabama AL       baldwin     01003       US PRESIDENT

US Presidential Election 2000-2020 Data (cont’d)

# last 6 columns
dat |> select(7:12) |> slice_head(n = 10)
# A tibble: 10 × 6
   candidate      party      candidatevotes totalvotes  version mode 
   <chr>          <chr>               <dbl>      <dbl>    <dbl> <chr>
 1 AL GORE        DEMOCRAT             4942      17208 20220315 TOTAL
 2 GEORGE W. BUSH REPUBLICAN          11993      17208 20220315 TOTAL
 3 RALPH NADER    GREEN                 160      17208 20220315 TOTAL
 4 OTHER          OTHER                 113      17208 20220315 TOTAL
 5 AL GORE        DEMOCRAT            13997      56480 20220315 TOTAL
 6 GEORGE W. BUSH REPUBLICAN          40872      56480 20220315 TOTAL
 7 RALPH NADER    GREEN                1033      56480 20220315 TOTAL
 8 OTHER          OTHER                 578      56480 20220315 TOTAL
 9 AL GORE        DEMOCRAT             5188      10395 20220315 TOTAL
10 GEORGE W. BUSH REPUBLICAN           5096      10395 20220315 TOTAL

2020 Presidential Election

Let’s focus on the 2020 Presidential Election

dat2020 = filter(dat, year == 2020)

dat2020 |> distinct(candidate)
# A tibble: 4 × 1
  candidate        
  <chr>            
1 JOSEPH R BIDEN JR
2 OTHER            
3 DONALD J TRUMP   
4 JO JORGENSEN

2020 Presidential Election, California Results

dat2020 |> 
  filter(state == "california") |>
  select(county_name, candidate:totalvotes) |>
  slice_head(n = 5)
# A tibble: 5 × 5
  county_name candidate         party       candidatevotes totalvotes
  <chr>       <chr>             <chr>                <dbl>      <dbl>
1 alameda     JOSEPH R BIDEN JR DEMOCRAT            617659     770070
2 alameda     OTHER             GREEN                 4664     770070
3 alameda     JO JORGENSEN      LIBERTARIAN           6295     770070
4 alameda     OTHER             OTHER                 5143     770070
5 alameda     DONALD J TRUMP    REPUBLICAN          136309     770070

Expressing votes relative to total in county

Let’s add a column propvotes to get the proportion of votes that each candidate obtained in every county:

propvotes = candidatevotes / totalvotes

votes_california = dat |>
  filter(state == "california" & year == 2020) |>
  mutate(propvotes = round(candidatevotes / totalvotes, 2)) |>
  select(county_name, candidate, propvotes, candidatevotes)

Analysis of California

Number of votes that each candidate received in each of the counties in California

votes_california |> 
  group_by(county_name, candidate) |>
  summarise(sum_votes = sum(candidatevotes))
# A tibble: 232 × 3
# Groups:   county_name [58]
   county_name candidate         sum_votes
   <chr>       <chr>                 <dbl>
 1 alameda     DONALD J TRUMP       136309
 2 alameda     JO JORGENSEN           6295
 3 alameda     JOSEPH R BIDEN JR    617659
 4 alameda     OTHER                  9807
 5 alpine      DONALD J TRUMP          244
 6 alpine      JO JORGENSEN             15
 7 alpine      JOSEPH R BIDEN JR       476
 8 alpine      OTHER                     6
 9 amador      DONALD J TRUMP        13585
10 amador      JO JORGENSEN            349
# ℹ 222 more rows

Maps

Map of US (contiguous states)

We’ve seen how to plot a map of US

# map data (as "sf" object)
us_states_sf = ne_states(
  country = "United States of America", 
  returnclass = "sf")

ggplot(data = us_states_sf) +
  geom_sf() +
  coord_sf(xlim = c(-125, -60), ylim = c(20, 55)) +
  theme_minimal()

Map of US (contiguous states)

Map of US Counties

"rnaturalearth" does not come with a built-in map data of US Counties. But we can use the "county" map-data from the package "maps".

To be consistent with the way we handle vector data, we convert the "county" map object into an "sf" object with st_as_sf().

us_counties_sf = st_as_sf(maps::map("county", plot = FALSE, fill = TRUE))

ggplot(data = us_counties_sf) +
  geom_sf() +
  coord_sf(xlim = c(-125, -60), ylim = c(20, 55)) +
  theme_minimal()

Map of US Counties

Map of California Counties

Map Data of California Counties

  • We have map-data of US states: us_states_sf

  • We have map-data of US counties: us_counties_sf

  • It would be nice to have map-data of California Counties.

How do get map-data of California Counties? This requires a bit of string matching via str_detect():

# matching "california" in ID column
cal_counties_sf = us_counties_sf |>
  filter(str_detect(ID, "^california"))

Map of California Counties

With filtered counties of California cal_counties_sf, we can make a map:

ggplot(data = cal_counties_sf) +
  geom_sf() +
  theme_minimal()

Map of California Counties

What’s in cal_counties_sf?

cal_counties_sf |> slice_head(n = 10)
Simple feature collection with 10 features and 1 field
Geometry type: MULTIPOLYGON
Dimension:     XY
Bounding box:  xmin: -124.2344 ymin: 35.90154 xmax: -118.3502 ymax: 42.00927
Geodetic CRS:  +proj=longlat +ellps=clrk66 +no_defs +type=crs
                        ID                           geom
1       california,alameda MULTIPOLYGON (((-121.4785 3...
2        california,alpine MULTIPOLYGON (((-120.0748 3...
3        california,amador MULTIPOLYGON (((-120.0748 3...
4         california,butte MULTIPOLYGON (((-121.6217 3...
5     california,calaveras MULTIPOLYGON (((-120.069 38...
6        california,colusa MULTIPOLYGON (((-121.8223 3...
7  california,contra costa MULTIPOLYGON (((-121.5702 3...
8     california,del norte MULTIPOLYGON (((-123.6844 4...
9     california,el dorado MULTIPOLYGON (((-121.1519 3...
10       california,fresno MULTIPOLYGON (((-120.6821 3...

Adding Column of County Name

For sake of convenience, we need to add a column county to the map-data cal_counties_sf (so that we have the name of the county)

cal_counties_sf = cal_counties_sf |>
  mutate(county = str_remove(ID, "california,"))

head(cal_counties_sf$county)
[1] "alameda"   "alpine"    "amador"    "butte"     "calaveras" "colusa"


View equivalent code 
cal_counties_sf$county = str_remove(
  cal_counties_sf$ID,
  "california,"
)

What relevant data do we have so far?

  • 2020 votes-data from California: votes_california

  • Map-data of California counties: cal_counties_sf

We need to join these data sets in order to combine the votes information with the map-data.

Joining map-data with elections-data

Join California map-data with votes-data, via inner_join()

# data cal_counties_sf has column 'county'
# data votes_california has column 'county_name'
cal_votes_map = inner_join(
  cal_counties_sf,
  votes_california, 
  by = c("county" = "county_name"))

Joining map-data with elections-data

cal_votes_map |> slice_head(n = 8)
Simple feature collection with 8 features and 5 fields
Geometry type: MULTIPOLYGON
Dimension:     XY
Bounding box:  xmin: -122.2749 ymin: 37.45998 xmax: -119.5362 ymax: 38.90956
Geodetic CRS:  +proj=longlat +ellps=clrk66 +no_defs +type=crs
                  ID  county         candidate propvotes candidatevotes
1 california,alameda alameda JOSEPH R BIDEN JR      0.80         617659
2 california,alameda alameda             OTHER      0.01           4664
3 california,alameda alameda      JO JORGENSEN      0.01           6295
4 california,alameda alameda             OTHER      0.01           5143
5 california,alameda alameda    DONALD J TRUMP      0.18         136309
6  california,alpine  alpine JOSEPH R BIDEN JR      0.64            476
7  california,alpine  alpine             OTHER      0.01              4
8  california,alpine  alpine      JO JORGENSEN      0.02             15
                            geom
1 MULTIPOLYGON (((-121.4785 3...
2 MULTIPOLYGON (((-121.4785 3...
3 MULTIPOLYGON (((-121.4785 3...
4 MULTIPOLYGON (((-121.4785 3...
5 MULTIPOLYGON (((-121.4785 3...
6 MULTIPOLYGON (((-120.0748 3...
7 MULTIPOLYGON (((-120.0748 3...
8 MULTIPOLYGON (((-120.0748 3...

Mapping Votes (facet by candidate)

View code 
ggplot(cal_votes_map) +
  geom_sf(aes(fill = propvotes)) +
  scale_fill_gradient(low = "#FFF7F5", high = "#FF0000") +
  facet_wrap(~ candidate) +
  theme_void()

Looking for an alternative color scheme

We can use a Viridis Color palette.

https://ggplot2.tidyverse.org/reference/scale_viridis.html


The function scale_fill_viridis_c() allows us to choose a continuous scale.

Its argument direction = -1 gives a reversing order (from yellow to dark blue).

Mapping Votes (5th attempt)

View code 
# Viridis Color
ggplot(cal_votes_map) +
  geom_sf(aes(fill = propvotes)) +
  scale_fill_viridis_c(name = "Prop. of Votes", direction = -1) +
  facet_wrap(~ candidate) +
  theme_void()

Map of Joe Biden’s votes

View code 
cal_votes_map |>
  filter(candidate == "JOSEPH R BIDEN JR") |>
ggplot() +
  geom_sf(aes(fill = propvotes)) +
  scale_fill_viridis_c(name = "Prop. of Votes", direction = -1) +
  theme_void() +
  labs(title = "Joe Biden's Proportion of Votes")

Joe Biden’s votes: top-5 counties

votes_california |>
  filter(candidate == "JOSEPH R BIDEN JR") |>
  arrange(desc(propvotes)) |>
  slice_head(n = 5)
# A tibble: 5 × 4
  county_name   candidate         propvotes candidatevotes
  <chr>         <chr>                 <dbl>          <dbl>
1 san francisco JOSEPH R BIDEN JR      0.85         378156
2 marin         JOSEPH R BIDEN JR      0.82         128288
3 alameda       JOSEPH R BIDEN JR      0.8          617659
4 santa cruz    JOSEPH R BIDEN JR      0.79         114246
5 san mateo     JOSEPH R BIDEN JR      0.78         291496

Joe Biden’s votes: bottom-5 counties

votes_california |>
  filter(candidate == "JOSEPH R BIDEN JR") |>
  arrange(propvotes) |>
  slice_head(n = 5)
# A tibble: 5 × 4
  county_name candidate         propvotes candidatevotes
  <chr>       <chr>                 <dbl>          <dbl>
1 lassen      JOSEPH R BIDEN JR      0.23           2799
2 modoc       JOSEPH R BIDEN JR      0.26           1150
3 tehama      JOSEPH R BIDEN JR      0.31           8911
4 shasta      JOSEPH R BIDEN JR      0.32          30000
5 glenn       JOSEPH R BIDEN JR      0.35           3995

Map with ggplotly()

View code 
joe_biden = cal_votes_map |>
  filter(candidate == "JOSEPH R BIDEN JR")

joe_biden_map = ggplot(data = joe_biden, aes(label = county)) +
  geom_sf(aes(fill = propvotes)) +
  scale_fill_viridis_c(name = "Prop. of Votes", direction = -1) +
  theme_void() +
  labs(title = "Joe Biden's Proportion of Votes")

ggplotly(joe_biden_map)

Return to Home Page