2018 Staten Island Election Results
January 21, 2019
Staten Island
Staten Island is the most conservative of the 5 New York City boroughs. In 2016, 56% of Staten Islanders voted for Donald Trump (compared to just 37% for all of NYC). Trump received 82% of possible Staten Island votes during the 2016 Republican Primary. If Staten Island was it’s own city, it would be the 2nd largest conservative city in America.
In 2018, Dan Donovan (R) was seeking reelection for New York’s 11th Congressional District. The 11th district combines all of Staten Island with some parts of southern Brooklyn (mostly Bay Ridge). Staten Island voters represent 3/4 of all possible congressional votes. Max Rose, the Democratic challenger did not stand much of a chance. fivethirthyeight.com gave Rose a 1 out of 4 chance to win. On the night of the election, predictit.org was paying out 3.70$ for every 1$ bet on Rose to win.
Rose ran a smart campaign. He was able to overcome the fact that he is not a native Staten Islander. He appealed to conservative voters by brandishing his military service record.
Max Rose ended up winning the election with 53% of the vote. In Staten Island, Rose received about half of all possible votes. Rose’s victory may have been the biggest congressional upset in 2018.
Plotting Staten Island Election Districts
New York City election district information can be found here. The shape file is also saved on github. Plotting shape files can be intimidating. This tutorial can explain any steps below that are glossed over below.
The rgdal library will be used to read the shape file. The fortify function from ggplot2 creates a data frame using the spatial data.
One thing to note before jumping in, the maptools library is dependent upon gpclib. Permission to use gpclib must be turned on to use maptools. Use gpclibPermit() to turn on the permission. This stack overflow article explains this issue more thoroughly.
library(rgdal) #readOGR
library(maptools)
library(ggplot2) #fortify
library(dplyr) #left_join
stat_island = readOGR("data", layer="geo_export", verbose=F)
stat_island@data$id = row.names(stat_island@data)
stat_island.points = fortify(stat_island, region="id")
stat_island.df = left_join(stat_island.points, stat_island@data, by="id")
ggplot(stat_island.df) + aes(long, lat, group=group) + geom_polygon()
2018 election information can be found here. A cleaned up data set can be found on github. After joining the Staten Island election results with the election district shape file, the map can be filtered for only election districts in Staten Island. This is also a good opportunity to change the theme of our plot using the ggthemes package.
library(ggthemes)
results18 = read.csv("data/election_results_2018.csv", stringsAsFactors = FALSE)
#join the election resluts
stat_island.df = left_join(stat_island.df, results18, by="elect_dist")
#filter for only districts where votes were recorded
stat_island.df = filter(stat_island.df, !is.na(TotalVotes))
ggplot(stat_island.df) +
aes(long, lat, group=group) +
geom_polygon(color="gray", fill="white") +
theme_void()
This next step is not mandatory but it helps the visual. There are three islands in the plot that are uninhabited and should be removed. The Staten Island Expressway is also added to show where the traditional North Shore boundary is.
#boxed long and lat of islands
stat_island.df$island_long =
ifelse(
(stat_island.df$long > -74.0558 & stat_island.df$long < -74.0466) |
(stat_island.df$long > -74.1625 & stat_island.df$long < -74.1570) , T, F)
stat_island.df$island_lat =
ifelse(
(stat_island.df$lat > 40.5645 & stat_island.df$lat < 40.5807) |
(stat_island.df$lat > 40.6411 & stat_island.df$lat < 40.6456) , T, F)
stat_island.df$island =
ifelse(stat_island.df$island_long & stat_island.df$island_lat, T, F)
#remove anything that is an island (except for staten island)
stat_island.df = filter(stat_island.df, island!=TRUE)
#staten island expressway coordinates
si_exp = read.csv("data/si_exp.csv")
ggplot(stat_island.df) +
aes(long, lat, group=group) +
geom_polygon(color="gray", fill="white") +
geom_smooth(data=si_exp, aes(x=long, y=lat, group=NULL), color="yellow", size=3) +
theme_void()
Election Results
The added shading depicts the winner of each district:
ggplot(data=stat_island.df) +
aes(long, lat, group=group) +
geom_polygon(aes(fill=stat_island.df$Winner), color="gray") +
scale_fill_manual(values=c("red", "blue")) +
geom_smooth(data=si_exp, aes(x=long, y=lat, group=NULL), color="yellow", size=3) +
theme_void() +
theme(legend.position="right", legend.title=element_blank())
Side by side plot to see where each candidate got their votes:
donovan = ggplot(data=stat_island.df) +
aes(long, lat, group=group) +
geom_polygon(aes(fill=stat_island.df$DD), color="gray") +
scale_fill_gradient(low="white", high="red") +
theme_void() +
theme(legend.position="bottom") +
labs(fill="Total Votes", title="Dan Donovan Votes")
rose = ggplot(data=stat_island.df) +
aes(long, lat, group=group) +
geom_polygon(aes(fill=stat_island.df$MR), color="gray") +
scale_fill_gradient(low="white", high="blue") +
theme_void() +
theme(legend.position="bottom") +
labs(fill="Total Votes", title="Max Rose Votes")
#use grid.arrange function from the gridExtra library
gridExtra::grid.arrange(donovan, rose, ncol=2)
Results from the districts where the most total votes were recorded:
stat_most = filter(stat_island.df, TotalVotes > 550)
#to differentiate between R and D win diff, D diff * -1
stat_most$plot_diff = ifelse(stat_most$Winner=="ROSE",
stat_most$Difference*-1,
stat_most$Difference)
ggplot(data=stat_most) +
aes(long, lat, group=group) +
geom_polygon(data=stat_island.df, fill="white", color="gray") +
geom_polygon(aes(fill=plot_diff, color=Winner)) +
scale_fill_gradient(low="blue", high="red") +
scale_color_manual(values=c("red", "blue")) +
theme_void() +
theme(legend.position="none") +
labs(title="Districts with at least 550 Total Votes Recorded")
North Shore Districts in the Red
Dan Donovan won 20 districts that are north of the Staten Island Expressway. Most of these districts are in Westerleigh or Castleton Corners, the two farthest east are in West Brighton. Donovan received 56% of the vote from these districts (compared to 48% of the votes from all other districts). These 20 districts make up 8% of all Staten Island votes. This is where the battleground for the 2020 election will be. Finding what these voters care most about and addressing it will be imperative for the winner of the next election.
#districts to hi-light in on
red_north = c(61034:61036, 61041:61043, 61045:61048,
63035, 63059, 63062:63067, 63072, 63081)
red_north.df = filter(stat_island.df, elect_dist %in% red_north)
red_north.plot = ggplot(data=red_north.df) +
aes(long, lat, group=group) +
geom_polygon(data=stat_island.df, fill="white", color="gray") +
geom_polygon(aes(fill=Difference), color="red") +
scale_fill_gradient(low="white", high="red", name="Margin of\nVictory\n",
breaks=c(.1,.2,.3), labels=c("10%","20%","30%")) +
theme_void() +
theme(legend.position="right")
red_north.plot
Taking a closer look:
red_north.plot +
coord_cartesian(xlim=c(-74.155, -74.105), ylim=c(40.605, 40.633)) +
theme(legend.position="none")
After reading this post from fivethirtyeight.com, I wanted to create similar plots using Staten Island voting info. Thanks a lot to Ella Koeze for creating the original plots. The originals were created using QGIS, I adapted them using R.
library(rgeos) #gCentroid function
#results from 2016 and 2018
results1618 = read.csv("data/results1618.csv")
#find the centroids using the original shapefile
stat_island$centroid = gCentroid(stat_island, byid=T)
#create a data frame
centroids = data.frame(
elect_dist=stat_island$elect_dist,
centroid.x=stat_island$centroid$x,
centroid.y=stat_island$centroid$y
)
#add election results to centroid df
centroids.df = centroids %>%
#filter for EDs of interest
filter(elect_dist %in% red_north) %>%
#join the data
left_join(results1618, by="elect_dist")
ggplot(centroids.df) +
geom_polygon(data=stat_island.df, aes(long, lat, group=group),
fill="white", color="gray80") +
geom_polygon(data=red_north.df, aes(long, lat, group=group),
fill="white", color="red") +
geom_point(aes(x=centroid.x, y=centroid.y, size=Vote.Share.Margin),
color="pink") +
scale_size_continuous(name="Margin of Victory",
breaks=c(.1,.2,.3,.4,.5),
labels=c("10%", "20%", "30%", "40%", "50%")) +
theme_void() +
theme(legend.position="bottom") +
coord_cartesian(xlim=c(-74.155, -74.105), ylim=c(40.605, 40.633)) +
facet_wrap(~Year, ncol=2)
In 2016, Donovan won all 20 of these districts just like in 2018. However, the margin of victory in these districts have been reduced drastically compared to 2016.