Visualization Essay: New Nation Votes
Alyssa Fahringer
library(readr)
library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionlibrary(stringr)
library(ggplot2)
library(tidyr)
url <- "http://dl.tufts.edu/file_assets/generic/tufts:MS115.003.001.00001/0"
if (!file.exists("all-votes.tsv")) {
download.file(url, "nnv-all-votes.zip")
unzip("nnv-all-votes.zip", files = "all-votes.tsv")
}
nnv <- read_tsv("all-votes.tsv")
names(nnv) <- names(nnv) %>%
str_to_lower() %>%
str_replace_all("\\.", "") %>%
str_replace_all("\\s", "_")
nnv <- nnv %>%
mutate(year = str_extract(date, "\\d{4}") %>% as.integer())Part One
A New Nation Votes is a dataset that is still in the process of being digitized and is the result of the efforts of Philip Lampi. The dataset consists of election results from the early Republic, broken down by state, year, office, jurisidiction, party, and election type. Currently, 18040 elections have been digitized, which is roughly 60% of Lampi’s overall collection. The election data was culled from contemporary sources, including newspapers and election ballots, in addition to nineteenth century county histories. The dataset is 107 MB, and contains 659, 625 observations of 25 variables.
summary(nnv)## id date type
## Length:659625 Length:659625 Length:659625
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## iteration office office_scope
## Length:659625 Length:659625 Length:659625
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## role role_scope state
## Length:659625 Length:659625 Length:659625
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## territory city county
## Length:659625 Length:659625 Length:659625
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## district town township
## Length:659625 Length:659625 Length:659625
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## ward parish populated_place
## Length:659625 Length:659625 Length:659625
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## hundred borough name
## Length:659625 Length:659625 Length:659625
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## name_id affiliation affiliation_id vote
## Length:659625 Length:659625 Length:659625 Min. : 0
## Class :character Class :character Class :character 1st Qu.: 16
## Mode :character Mode :character Mode :character Median : 64
## Mean : 305
## 3rd Qu.: 186
## Max. :128519
## NA's :60996
## year
## Min. :1787
## 1st Qu.:1802
## Median :1810
## Mean :1810
## 3rd Qu.:1819
## Max. :1826
## What kinds of elections were there, and how many of each kind of election?
nnv %>% count(type)## Source: local data frame [6 x 2]
##
## type n
## (chr) (int)
## 1 General 620179
## 2 Legislastive 11
## 3 Legislative 24576
## 4 Special 14770
## 5 Special Election 14
## 6 Special Legislative 75nnv %>%
ggplot(aes(x = type)) +
geom_bar(stat = "count")
Using the count function, it is evident that there are six types of elections represented in the dataset: general, legislative, special, legislastive, special legislative, and special election. I am going to assume that ‘legislastive’ is a spelling error. I am also curious as to the difference between ‘special’ and ‘special election.’ The overwhelming majority of the elections were general elections.
How many candidates are represented and how often do they appear?
nnv %>%
count(name, name_id) %>%
ungroup() %>%
arrange(desc(n))## Source: local data frame [41,738 x 3]
##
## name name_id n
## (chr) (chr) (int)
## 1 Caleb Strong SC0023 6282
## 2 William Phillips PW0081 6136
## 3 Elbridge Gerry GE0049 5420
## 4 others null 3993
## 5 James Sullivan SJ0366 3930
## 6 William Heath HW0171 3910
## 7 Edward H. Robbins RE0018 3693
## 8 scattering null 3650
## 9 Samuel Adams AS0022 3390
## 10 Moses Gill GM0022 3298
## .. ... ... ...Using the count, ungroup, and arrange functions, there are over 41 thousand candidates represented in the dataset. Caleb Strong is the candidate who is most represented in the dataset, as he appaears over six thousand times.
Which parties?
nnv_aff <- nnv %>%
filter(affiliation != "null")
nnv_aff %>% count(affiliation)## Source: local data frame [178 x 2]
##
## affiliation n
## (chr) (int)
## 1 1st Ticket Democratic Republican 80
## 2 2nd Ticket Democratic Republican 80
## 3 Adamite 4
## 4 Administration 24
## 5 Administration Ticket 8
## 6 Against New Election 80
## 7 American 40
## 8 American Republican 102
## 9 Anit-Republican 6
## 10 Anti Federal 32
## .. ... ...Using the affiliation variable, there are 178 affiliations, discounting null and NA.
What does each row in the dataset represent?
Each row is a particular candidate running in a particular election in a particular year, and sometimes in a specific region.
Which years are in the dataset, and how many elections are there in each year?
nnv <- nnv %>%
mutate(year = str_extract(date, "\\d{4}") %>% as.integer())
nnv %>%
group_by(year) %>%
summarise(unique_elements = n_distinct(id))## Source: local data frame [40 x 2]
##
## year unique_elements
## (int) (int)
## 1 1787 69
## 2 1788 159
## 3 1789 186
## 4 1790 168
## 5 1791 156
## 6 1792 194
## 7 1793 167
## 8 1794 243
## 9 1795 202
## 10 1796 357
## .. ... ...Within the dataset, the years 1787 through 1826 are represented. 1824 had the most elections out of those years, topping out at 1221. The final year of the dataset, 1826, had the lowest number of elections at 6.
Which states are represented?
nnv_state <- nnv %>%
filter(state != "null")
nnv_state %>% count(state)## Source: local data frame [25 x 2]
##
## state n
## (chr) (int)
## 1 Alabama 2422
## 2 Connecticut 5250
## 3 Delaware 10224
## 4 Georgia 8490
## 5 Illinois 4805
## 6 Indiana 4084
## 7 Kentucky 8951
## 8 Louisiana 1610
## 9 Maine 10840
## 10 Maryland 29839
## .. ... ...There are 25 states represented, discounting NA and null. NY had the least number of elections–only ten–out of all the states represented, although there is a great possibility that this was a mistake on the part of the researcher who input the data, as there is also data for New York. The state with the highest number of elections is Massachusetts, which had 157732 elections.
Part Two
nnv_loudoun <- nnv %>%
filter(state == "Virginia", county == "Loudoun", office == "U.S. House of Representatives")
nnv_loudoun %>% count(name) ## Source: local data frame [21 x 2]
##
## name n
## (chr) (int)
## 1 Armistead T. Mason 2
## 2 Arthur Lee 1
## 3 Bushrod 1
## 4 Charles F. Mercer 5
## 5 John Littlejohn 1
## 6 John Love 2
## 7 John Pope 1
## 8 John Washington 1
## 9 Joseph Lane 1
## 10 Joseph Lewis 2
## .. ... ...nnv_loudoun <- nnv_loudoun %>%
filter(affiliation != "null")
nnv_loudoun %>%
ggplot(aes(x = affiliation)) +
geom_bar(stat = "count") +
labs(title = "number of votes cast by party in Loudoun",
x = "party affiliation",
y = "number of votes")
nnv_loudoun %>%
group_by(affiliation, year) %>%
summarize(vote = sum(vote, na.rm = TRUE)) %>%
ggplot(aes(x = year, y = vote, color = affiliation)) +
geom_line() +
labs(title = "total votes per year in Loudoun County",
x = "year",
y = "number of votes")
I wanted to examine data from my home county of Loudoun.Between the years of 1797 and 1825, 21 people ran for the US House of Representatives in Virginia. Because I wanted to examine the number of votes placed per party, I filtered out all votes where the affiliation was listed as null. The amount of votes placed for Federalist and Republican candidates were almost equal, although Federalist candidates ultimately received the most votes.
nnv_joseph <- nnv_loudoun %>%
filter(name_id == "LJ0565")
nnv_joseph %>%
ggplot(aes(x = year, y = vote)) +
geom_line() +
labs(title = "votes for Joseph Lewis, Jr.",
x = "year",
y = "number of votes")
Going through the data in the nnv_loudoun dataset, I noticed that one person, Joseph Lewis/Joseph Lewis, Jr. ran for the House of Representatives multiple years. I knew it was the same person because the name id was the same for both Joseph Lewis and Joseph Lewis, Jr. The year he received the most votes was 1815, and he received the least votes in 1811.
ny_gov <- nnv %>%
filter(name_id != "null", office == "Governor", state == "New York", is.na(district), is.na(county), is.na(city)) %>%
distinct(year, name_id) %>%
select(name, name_id, vote, year, affiliation) ggplot(ny_gov, (aes(x = name, y = vote))) +
geom_bar(stat = "identity") +
labs(title = "Votes for governor in New York",
x = "year",
y = "number of votes")## Warning: Removed 2 rows containing missing values (position_stack).
ny_gov %>%
filter(year == "1816") %>%
ggplot(aes(x = name, y = vote, fill = affiliation)) +
geom_bar(stat = "identity") +
labs(title = "votes for Governor of New York in 1816",
x = "name of candidate",
y = "number of votes")
I wanted to examine a state that had a large number governor observations, which led me to New York. I whittled down the data to show the state-wide votes for each candidate for governor for each year. I grouped by year and had the data arranged in descending order by vote. I used a facet wrapped graph to show the I visualized the data in a facet wrap graph. Because the labels along the X axis were distorted, this graph was not as helpful as it could have been. I then chose a year that had a relatively high number of votes to look more closely at one election year, since the facet wrap did not display in a readable way.