University of Southern California Dissertations and Theses

Wake County District Overlay: an online electoral data visualization application

(USC Thesis Other)

Wake County District Overlay: an online electoral data visualization application

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content WAKE COUNTY DISTRICT OVERLAY:

AN ONLINE ELECTORAL DATA VISUALIZATION APPLICATION

by

Haynes Hoyle Bunn

A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

December 2014

Copyright 2014 Haynes Hoyle Bunn

ii

ACKNOWLEDGMENTS
There are many people who have helped me during my Master’s program, not only in
working on this thesis project but also in encouraging me throughout these past five semesters.
I’d especially like to thank my mom and dad who cheered me on as I worked towards this latest
educational goal, and helped me immeasurably, both financially and mentally. I am forever
grateful. I’d also like to thank the rest of my family, who may be too numerous to list here
individually, but should still be recognized. I know I haven’t seen many of you in a long time,
but know that once my Master’s is finished I will finally have time to visit!
I’d also like to thank several people who have helped specifically with this project. Of
course, my thesis committee, Dr. Jennifer Swift, Dr. Daniel Warshawsky, and Committee Chair
Dr. Yao-Yi Chiang, without whom I would not meet this critical deadline; Dr. Chiang has been
an incredible help throughout the process. I’d also like to thank Ed Shipman, who empathizes
with my midnight crunch-time web coding. I’d like to thank all of the Gospodareks, who have
helped me immensely in my academic career. And I’d like to thank all of my colleagues at
America Votes who tested my application and gave me great feedback, and my USC colleagues
who helped along the way.
Finally, I’d like to thank my friends who I haven’t seen “due to thesis” in many months
now. And I’d like to give a special shout out to my best friend Jillian Tomlinson and her syntax
repair shop.
And of course, I’d like to thank Rob Strong. Thank you for putting up with me always
having my nose in a textbook for these last 2 years. Now, let’s go on a vacation sans homework!

i

TABLE OF CONTENTS
ACKNOWLEDGMENTS ii
LIST OF TABLES iii
LIST OF FIGURES iv
LIST OF ABBREVIATIONS vi
ABSTRACT vii
CHAPTER 1: INTRODUCTION 1
1.1 Motivation 1
1.2 Web Application Overview 3
1.3 Choice of Technologies 5
CHAPTER 2: BACKGROUND AND LITERATURE REVIEW 7
CHAPTER 3: METHODOLOGY 13
3.1 Workflow for Wake County District Overlay Creation 13
3.1.1 Download Data 14
3.1.2 Create Database Diagram 17
3.1.3 Format Data 19
3.1.4 Create Database in PostGIS 21
3.1.5 Upload Database to GeoServer 23
3.1.6 Build Web Application Using Leaflet 26
3.2 CartoDB: Alternative to PostGIS/GeoServer/Leaflet for Web Application Design 33
CHAPTER 4: EVALUATION 37
4.1 Survey Results on Evaluator Experience 37

ii

4.2 Survey Results on Gauging Whether the Wake County District Overlay Meets the
Intended Purpose 40
4.3 Survey Results from Open-Ended Questions 43
CHAPTER 5: DISCUSSION AND CONCLUSION 45
5.1 Results of the Wake County District Overlay Web Application 45
5.2 Future Improvements to the Wake County District Overlay 46
REFERENCES 50
APPENDIX A: PYTHON SCRIPT FOR REARRANGING DATA IN EXCEL 54
APPENDIX B: GOOGLE EVALUATION SURVEY 55

iii

LIST OF TABLES
Table 1 Data Downloaded for Use in the Web Application 16

iv

LIST OF FIGURES
Figure 1 House District 34 (yellow) in Wake County, North Carolina 4
Figure 2 Workflow for Wake County District Overlay Creation 14
Figure 3 Database Diagram 18
Figure 4 Create Address Locator 20
Figure 5 PostGIS Database 22
Figure 6 Set up Workspace in GeoServer 23
Figure 7 GeoServer Layers 24
Figure 8 Example SQL View Layer 25
Figure 9 Web Map Showing Mapbox Basemap 26
Figure 10 Web Map Showing US Congressional Districts in Wake County and All Democratic
and Unaffiliated Voters in HD34 Displayed in the Wake County District Overlay Web
Application 27
Figure 11 Web Map Showing US Congressional Districts, State House Districts, State Senate
Districts in Wake County and All Democratic and Unaffiliated Voters in HD34 28
Figure 12 Web Map Showing US Congressional Districts and Voters in HD34 who voted in the
2008 general election, but did not vote in the 2012 general election 29
Figure 13 Web Map Showing US Congressional Districts and Voters in HD34 divided between
Congressional District 4 (purple) and Congressional District 13 (green) 30
Figure 14 Web Map Showing State House Districts, Precincts, and Voters in HD34 who have
ever voted using One-Stop Early Voting 31
Figure 15 Web Map Showing FCC Television Media Market Coverage in Wake County 32
Figure 16 Web Map Showing Nationwide FCC Television Media Market Coverage 33

v

Figure 17 CartoDB Web Map Alternative to Leaflet Map 35
Figure 18 Graph of Survey Question #4 Answers 38
Figure 19 Graph of Survey Question #5 Answers 39
Figure 20 Graph of Survey Question #6 Answers 40
Figure 21 Graph of Survey Question #1 Answers 41
Figure 22 Graph of Survey Question #2 Answers 42
Figure 23 Graph of Survey Question #3 Answers 42

vi

LIST OF ABBREVIATIONS
API Application Programming Interface
AV America Votes
CSS Cascading Style Sheets
FCC Federal Communications Commission
GIST Geographic Information Science and Technology
HD34 House District 34
SQL Structured Query Language
SRID Spatial Reference System Identifier
USC University of Southern California

vii

ABSTRACT
Political campaigns are inherently geography-driven. The creation of a tool that allows a non-
GIS (geographic information systems) professional to quickly and easily display the various
spatial data concerning voters and their locations is key to winning elections. A political non-
profit organization headquartered in Washington, D.C., has expressed an interest in a web-based
mapping application that aids in voter targeting, especially the targeting of registered voters with
irregular voting habits in order to persuade them to vote on Election Day. This thesis presents a
web-based application, which provides campaigns and organizations with fast access to the
knowledge they need to manage field operations. This project relies on open source software,
including Leaflet, PostGIS, and GeoServer. The project focuses on Wake County in North
Carolina, with the expectation of expanding the web application to the state level in the future.
The data needed for the project are readily available and include publicly available voter files of
all registered voters, and shapefiles of Wake County media markets, precincts, polling places,
state House and Senate districts, Congressional districts, School Board districts, County
Commissioner districts, and Superior Court districts. The evaluation shows that campaign
staffers can use the web application to efficiently and effectively visualize relevant combinations
of the above data and share this knowledge with their colleagues.

1

1

CHAPTER 1: INTRODUCTION
The use of Geographic Information Systems (GIS) in political campaigns has the potential to
make the campaign process more efficient by introducing increasingly accurate spatial data and
analysis. Because time is a valuable resource on a campaign, making this process as quick and
easy as possible is crucial. When attempting to maximize voter turnout, finding the persuadable
voters is key to winning the election. Finding the low turnout voters and persuading them to vote
in your favor can be the difference between a loss and a victory on Election Day. However,
many campaign staffers are not trained in GIS, therefore campaigns often do not have the
resources to display this spatial data visually.
Campaigns have access to data about voters, including their addresses and vote history,
but may not have the capacity to transform this from a tabular form into a map. The creation of a
web-based application streamlines the process of visually displaying data necessary for a
political campaign and allows for easy sharing between staffers and volunteers. As most
campaign staffers are not trained in GIS, the interactive web map is designed to be extremely
user-friendly. In addition to housing the data detailed below, it also allows for easy updating and
editing of information, as well as the addition of new data as it becomes available.
1.1 Motivation
Political campaigns are inherently geography-driven, so the creation of a web application
that allows a non-GIS professional to quickly and easily display the spatial data detailed below is
an integral part of optimizing the work of field staffers and volunteers. This thesis project blends
campaigning with spatial science, and builds a better visualization tool for targeting at the level
of the individual voter.
2

2

This web application provides campaigns and organizations with fast access to the
knowledge the organizations need to target voters in order to win the vote on Election Day. By
incorporating voter participation, election history, and redistricting data, as explained in section
3.1.1, this thesis project provides the latest information available to campaign staff and
volunteers.
Each election cycle brings a new technology to the forefront – voter targeting as it
currently stands has only existed for the last decade – and this web application continues that
evolution. Technological changes in campaign work were the stimulus for voter files
1
becoming
an integral tool in voter targeting in the first place (Blaemire 2013). The Help America Vote Act
of 2002, a result of the anomalies of the 2000 Presidential election, requires voter files of all
registered voters to be made publicly available by states, or subdivision of states (Schlozman
2010).
I am the Executive Assistant at America Votes (AV),
2
a political non-profit
headquartered in Washington, D.C. I serve as the assistant to both the President and the
Managing Director of the organization. AV works in 11 core states and 9 affiliate states on issue
advocacy and electoral politics, working to advance progressive policies and expand voting
rights. The organization provides data and targeting services to progressive organizations within
these 20 states, and works to provide coordination and collaboration among them. This
coordination not only helps to eliminate duplicative efforts by partner organizations in the states,
but also serves as the institutional memory between elections, creating a permanent campaign
infrastructure.

1
Voter files are publicly available data about registered voters, including voter registration number, address,
birthday, and past voting history.
2
www.americavotes.org
3

3

America Votes has expressed an interest in a web-based data visualization application
that overlays voter file information with political districts and other data that may aid in turning
out voters and winning elections. Theo Luebke, North Carolina Field Director for AV, requested
such an application in an email on March 28, 2014. In his job as AV-North Carolina Field
Director, he would use the Wake County District Overlay application to generate maps for
partners and activists; these maps would be the foundation for campaign decision-making
(Luebke 2014).
Additionally, in a conversation on April 2, 2014 with Christopher Gill, Data & Targeting
Director at America Votes, Mr. Gill indicated that the organization has been looking into
creating an application like the one presented in this thesis project. He observed that while maps
make an excellent communication vehicle, the organization’s ability to create them is limited to
its technical staffers, and that ideally mapping tools could be put in the hands of all staff.
Additionally, he suggested that should this Wake County project be implemented at the state or
national level, the organization would be interested in utilizing it for state-level work across the
country (Gill 2014).
1.2 Web Application Overview
The application developed as part of this thesis project focuses on Wake County in North
Carolina, and uses House District 34 as a case study. I grew up in Raleigh, the capital of North
Carolina, located in Wake County. Wake County has a population of about 1 million residents,
which increases by approximately 62 people per day (WakeGOV People and Places). The county
is known for having accessible public data, including GIS files and voter information. The
county makes voters files available online free of charge, and updates these voter files weekly.
These files include the name, address, voter registration number, age, race, and vote history
4

4

information for every registered voter in the county, and are detailed in subsequent sections of
this paper. Figure 1 shows the location of the House District within Wake County in North
Carolina.

Figure 1 House District 34 (yellow) in Wake County, North Carolina

The final product is a web-based application that can be easily used by a campaign or
organization within Wake County. The application aids in determining where to find low
propensity voters and helps the user determine how best to reach them. The Wake County
District Overlay application is interactive, allowing the user to display data on a map and overlay
5

5

relevant layers, including various political districts, precincts, polling places, and voter history.
The project requires the use of publicly available voter files of all registered voters, including
past vote history, to determine registered voters who seldom vote or may be inactive. As the
Wake County voter files are updated weekly on the Wake County Board of Elections website,
the voter file from the week of June 27, 2014 is used is in this thesis project. At the time the
addresses from the voter file for this case study were geocoded (detailed in Chapter 3), this was
the most up-to-date voter file available. Additionally, the shapefiles, or files that store geometry
and attribute information for display in a GIS, of the various political districts within Wake
County are included. (Esri 1998). Shapefiles of Federal Communications Commission media
markets from the FCC website, and precincts, state House and Senate districts, Congressional
districts, School Board districts, County Commissioner districts, and Superior Court districts
from the Wake County Map Services website are included in the web application.
1.3 Choice of Technologies
To build the web application of this thesis, I installed the OpenGeo Suite, an open source
geospatial platform, on a 2011 MacBook Pro laptop. OpenGeo includes an instance of PostGIS,
an open source spatial database, and GeoServer, an open source server for web pages that display
spatial data (The PostGIS Development Group; Geoserver). For future development, the
OpenGeo Suite can be installed on a server to allow for a larger data set and improve data
accessing time. Finally, Leaflet, an open source JavaScript
3
mapping library, is used to build and
style the website (Leaflet). Additionally, a prototype of a more interactive map than can be
supplied by GeoServer is built on CartoDB, a cloud-based mapping tool, for demonstration

3
A web programming language.
6

6

purposes, but is not the core subject of the thesis project (CartoDB). This is described in more
detail in Chapter 3.
Esri products, and other proprietary GIS software, are not used as the foundation of this
application. As is discussed in Chapter 3, ArcMap is used to geocode the addresses, but the
database and web platform use open source tools. America Votes does not use Esri products, so
in order to implement this application beyond the scope of the thesis project, the application uses
tools that are available to the organization.

7

7

CHAPTER 2: BACKGROUND AND LITERATURE REVIEW
When determining which voters to target, one must first have an understanding of the electorate.
In general, the voters on the far end of the ideological spectrum matching one’s campaign
already support the candidate and do not require outreach from the campaign to turn out on
Election Day. Likewise, the voters on the opposite end of the ideological spectrum likely do not
support the candidate. Therefore time should not be spent on these voters either. It is the voters
in the middle of the spectrum who determine the outcome of elections, but need to be persuaded
before Election Day. These voters need to hear messages from the campaign that speak to the
issues they care about, and they need to hear these messages many times over (Keschmann 2013
and Williams 2014).
The best way for campaigns to make sure voters hear these persuasion messages is
through one-on-one conversations. However, on a campaign there is limited time to have these
conversations, so staffers must prioritize which voters to reach (Keschmann 2013). For example,
in the 2006 gubernatorial campaign in Texas, Rick Perry’s campaign focused on electoral
targeting; however they were focused at the county level, rather than at the level of the individual
voter (Cho and Gimpel 2009). The campaign knew they needed to prioritize the voters they
spoke with, but did not target at the level of the individual voter. Instead, the campaign decided
to take a macro-level approach and canvass based on county population and density. This
allowed the campaign to maximize the staff and volunteer hours by reaching as many voters as
possible while avoiding the sparsely populated areas. But as Darmofal found in his 2006 article,
groups of voters tend to act and vote differently than expected at the individual level (Darmofal
2006). Therefore, targeting voters in these larger blocks, as the Perry campaign did with counties
in Texas, is often not as worthwhile as targeting individual voters.
8

8

A tool that enables this geographical targeting at this micro-level is necessary to improve
upon the process. Until the early 2000s, the accuracy of voter data at the individual level was
unreliable, and therefore was often not utilized within GIS programs (Sui and Hugill 2002).
Furthermore, the GIS programs themselves were often incapable of handling such vast amounts
of data (Sui and Hugill 2002). In their 2002 article, Sui and Hugill conclude that advancements
in these technologies will be the catalyst for increased use in campaign targeting and analysis.
Indeed, as political technology progresses, campaigns are moving from the “knock on every
door” tactic to the “knock on the doors that matter” strategy. This strategy of targeting individual
voters is called “microtargeting” (Altom 2012). In order to implement the microtargeting
strategy, campaigns must have researched and collected vast amounts of information about
voters, and must store this information in a database (Keschmann 2013). There are two databases
that are widely used by political campaigns and organizations: Voter Vault
4
on the conservative
side and Votebuilder,
5
also referred to as VAN, on the progressive side (Altom 2012). Both
databases allow for the storage of past voter contact history, which allows the campaigns to make
highly informed decisions regarding future voter contact. As these databases are proprietary
tools, there is not much public documentation or analysis surrounding the use and effectiveness
of the databases.
This lack of public access to proprietary campaign technology or datasets is
commonplace in the political sphere. For example, shortly after the 2012 Presidential election,
the technology staff for President Obama’s reelection campaign pushed to release the campaign’s
voter targeting technology back to the public so that outside organizations could continue to
build upon the foundation (Popper 2013). However, the campaign and the Democratic Party did

4
http://www.filpac.com/votervault.htm
5
http://www.votebuilder.com
9

9

not want to make the code public, claiming that it was proprietary information that could be used
against them by the Republican Party (Popper 2013). This inherent secrecy in political
campaigns contributes to the sparseness of research available on technological political tools.
The storage of voter information allows the campaigns to connect the persuadable voters
with the messages that speak to them (Keschmann 2013). For instance, if during a phone call in
2008 a voter indicated that the environment was an important issue, the 2012 campaign was able
to use that information to send that person a flyer about the candidate’s environmental policies.
This ensures that each voter receives information that is more likely to persuade him or her to
vote for the candidate. As voters receive more political information, the more likely they are to
seek out additional information and share their opinions (Cho 2011). This leads political
campaigns and organizations to actively seek out potential voters, hoping to persuade them 1) to
vote and 2) to vote a certain way on Election Day. American political parties have proven
themselves especially adept at reaching out to voters and turning them out to vote (Karp 2008).
And when parties, candidates, and organizations do turn out these voters, they want to be sure to
increase the level of information available so that they vote in their favor (Jessee 2010).
Storing voter information in a database is only the beginning of the microtargeting
process. Displaying that information visually is the next step, allowing the staffers to gain a
sense of the spatial data associated with the voter data. The use of GIS enables this spatial data
visualization for political campaigns, according to Cho (2012) instead of viewing each voter
individually. Displaying voters on a map allows for the collective trends to become evident.
In addition to reaching voters by mail, phone, and door-to-door, the campaigns can use
the Internet as a useful tool for delivering a campaign message (Kacor 2012). According to Jake
Williams’ 2014 article in Campaigns & Elections, “2013 was the first year that Americans spent
10

10

more time online than they did in front of a television” (Williams 2014). Campaigns must be
able to adapt to the changing habits of voters; it is crucial that campaigns embrace tools offered
by the Internet. The fact that most young voters are also Internet users allows campaigns to target
those voters specifically through social media advertisements, helping to boost the turnout in an
age bracket that often sees low turnout numbers (Schlozman 2010).
Williams goes on to describe four voter outreach strategies that campaigns should use in
2014, three of which are Internet-based. The first Internet outreach tactic is the utilization of
cookies in one-on-one targeting (Williams 2014). Cookies are pieces of information that are
stored in a user’s browser history when visiting a website. Third parties can then use this
information to track the user’s history and tailor website ads to the interest of that user. By using
cookies that are a part of a user’s web browser history, campaigns can determine exactly which
voters to reach out to in order to maximize the return on their investment. However, many
analysts caution against using the cookie method as the sole outreach strategy (Delany 2012).
Second, Williams urges the use of Facebook Custom Audiences, which allows a
campaign to match their existing email list with Facebook users. Instead of sending an email to
that list, the campaign could have a message, photo, or video show up in the Facebook newsfeed
of those voters (Williams 2014 and Delany 2011). Additionally, Facebook has recently
announced the addition of Managed Custom Audiences, which allows the campaign to target
those ads in both a desktop and mobile platform, maximizing their outreach potential (Williams
2014).
The third and final Internet outreach method suggested by Williams is Twitter Tailored
Audiences (Williams 2014). This allows a campaign to purchase advertising space in the feeds of
users who have already expressed an interest in that campaign via Twitter. Williams does point
11

11

out, however, that this method is not aimed at campaigns, but rather at companies who are
selling products (Williams 2014). There is potential for the growth of Twitter Tailored
Audiences in the political sphere, especially if expanded to include the use of the location
services within the Twitter API to ensure campaigns reach voters who live within the correct
districts.
Williams also asserts that set-top box data mining will be rampant in the 2014 election
cycle (Williams 2014). This allows a campaign to glean television-viewing habits of specific
locations in order to make informed television advertising decisions. Cable television ad buys
can be made within a “cable zone” to focus in on a specific geographic area, as EMILY’s List
did in a 2012 US House race in Arizona (White and Nuckels 2013). This strategy was made
possible by the ever-increasing access to numerous cable channels throughout the country
(Lilienthal 2005).
Studies of the spatial distribution of voters have also determined that voters who hold the
same policy opinions tend to live in the same areas, creating small enclaves of like-mindedness
(McKee 2009). Indeed, people in general tend to surround themselves with people who think and
act like them, creating a homogenous social environment (Cho 2011). This phenomenon works
in favor of the political entities that are conducting in-person persuasion campaigns, allowing for
microtargeting within smaller geographies.
To geographically target voters, campaign staff must be able to perform analysis and
visualize data for field workers. Maintaining data in a format that can be easily understood and
manipulated by campaign professionals is a priority, as campaign staffers are often not trained in
the use of GIS software (Rigamer 1998). And as Sheehan points out, the demand for interactive
web maps, especially those that can be optimized for mobile viewing, is increasing drastically
12

12

(Sheehan 2012). This is where the Wake County District Overlay project joins the microtargeting
culture of campaigns. The application created as part of this thesis project maintains spatial data
about voters at the individual level in a format that can be used by any campaign or organization
staff. The user can choose to display voters based on vote history, demographics, location within
political districts, and voting method, and can choose to display specific political district
boundaries. The web application can quickly display the relevant spatial data for determining
where to find the persuadable voters and how best to reach out to them, given their vote history
and social environment. Once hosted online, the web application can be shared among America
Votes staffers and volunteers throughout the 11 core states and 9 affiliate states, creating easy
accessibility for all users.

13

13

CHAPTER 3: METHODOLOGY
This chapter reviews the methodology behind the creation of the Wake County District Overlay
project, discusses the data sources used in the display, and provides a conceptual look at the
strengths and drawbacks of the open source tools driving the application. Section 3.1 discusses
the workflow for creating the Wake County District Overlay, utilizing PostGIS, GeoServer, and
Leaflet to complete the application. Section 3.2 examines CartoDB as an alternative method for
creating this application, and explains how the PostGIS/GeoServer/Leaflet version better meets
the intended purpose.
3.1 Workflow for Wake County District Overlay Creation
This section discusses the various steps of designing and building the Wake County
District Overlay web application. Figure 2 depicts the workflow, illustrating each design and
development segment of the project. The process begins by downloading all of the necessary
data and cataloguing all of the various attributes. Next, the database diagram is created. This is a
critical step to ensure the inclusion of all data and the respective relationships, and to determine
which attributes may be unnecessary for this project. Once the database design is determined,
any formatting changes necessitated by the design diagram are made, including deleting any
unnecessary columns and changing the column headings. The formatted data is then uploaded to
PostGIS to create the database and the relationships between the tables. After the creation of the
database, the data is then uploaded to GeoServer so that the spatial data can be published through
the web application. Finally, the web application is created using Leaflet, an open source tool
based in JavaScript. It is after this step that evaluation of the application begins.
14

14

Figure 2 Workflow for Wake County District Overlay Creation

3.1.1 Download Data
The first step in the process is to download all of the data necessary for display in the
Wake County District Overlay application. This project uses four groups of data: political district
boundaries, voter information, FCC media market boundaries and points, and a streets shapefile.
The political data includes shapefiles for the US Congressional Districts within Wake
County, state House Districts, state Senate Districts, Wake County Commissioner Districts,
Wake County School Board Districts, Wake County Superior Court Districts, precincts, and
3.1.1
Download
Data
3.1.2
Create
Database
Diagram
3.1.3
Format Data
Geocode
Addresses
Create
Address
Locator
Wake
County
Streets
Shapefile

Use ArcGIS
Geocoding
Tool
Voter File
Spreadsheet

Create
Vote_History
Table
Use
Python
Script to
Reformat
3.1.4
Create
Database in
PostGIS
3.1.5
Upload
Database to
GeoServer
3.1.6
Build Web
Applica on
using Leaflet
Wake County
District
Overlay Web
Applica on
15

15

polling places (Wake County Map Services). Wake County has an extensive online library of
GIS data for the county, which includes all of the shapefiles needed for this project.
The second group comprises three tables created from the publicly available voter files: a
shapefile of the geocoded addresses, a table of voter history, organized by Voter Registration
Number, and a vote history key (Wake County Voter Data). The shapefile allows the map to
display each registered voter downloaded in the voter file. The vote history table indicates which
elections voters have or have not voted in, and the vote history key indicates the voting method
for each voter in each election (e.g in-person, absentee, etc).
The third group is also composed of two data sets, including the FCC media markets
point shapefile and the media markets coverage polygon shapefile (Federal Communications
Commission GIS). The point shapefile displays the location of television station transmitters
throughout the country. The coverage polygon shapefile displays the area within which the
television signal should be attainable. These two shapefiles are downloaded from the website of
the Federal Communications Commission. This data is used in the Wake County District
Overlay Application in order to show whether voters are within range of television signals,
which can aid in purchasing political television advertisements.
And finally, the last piece of data is the Wake County streets shapefile, which is used as
the Address Locator in the geocoder used to geocode the addresses from the voter file (Wake
County Map Services). However, this shapefile is not used for display in the final web
application. Instead, a basemap is used, which includes the streets. Additional data information is
available in Table 1.

16

16

Table 1 Data Downloaded for Use in the Web Application
Name File Creation
Date
Source Type
Wake County
Congressional
Districts
March 2013 Wake County Map Services:
http://www.wakegov.com/gis
/services/Pages/data.aspx
Shapefile
Wake County State
House Districts
March 2013 Wake County Map Services:
http://www.wakegov.com/gis
/services/Pages/data.aspx
Shapefile
Wake County State
Senate Districts
March 2013 Wake County Map Services:
http://www.wakegov.com/gis
/services/Pages/data.aspx
Shapefile

Wake County
Precincts
November 2013 Wake County Map Services:
http://www.wakegov.com/gis
/services/Pages/data.aspx
Shapefile
Wake County
Polling Places
March 2014 Wake County Map Services:
http://www.wakegov.com/gis
/services/Pages/data.aspx
Shapefile
Wake County
Streets
May 2014 Wake County Map Services:
http://www.wakegov.com/gis
/services/Pages/data.aspx
Shapefile
FCC Media Markets
(points)
June 2012 Federal Communications
Commission GIS:
http://wireless.fcc.gov/geogr
aphic/index.htm?job=licensi
ng_database_extracts
Shapefile
FCC Media Markets
(coverage)
June 2012 Federal Communications
Commission GIS:
http://wireless.fcc.gov/geogr
aphic/index.htm?job=licensi
ng_database_extracts
Shapefile
Voter File (House
District 34
Democrats and
Unaffiliated Voters)
June 27, 2014 Wake County Voter Data:
http://www.wakegov.com/ele
ctions/data/pages/data.aspx
Spreadsheet
Wake County
Commissioners
Districts
October 2011 Wake County Map Services:
http://www.wakegov.com/gis
/services/Pages/data.aspx
Shapefile
Wake County
School Board
Districts
March 2013 Wake County Map Services:
http://www.wakegov.com/gis
/services/Pages/data.aspx
Shapefile
Wake County
Superior Court
Districts
November 2011 Wake County Map Services:
http://www.wakegov.com/gis
/services/Pages/data.aspx
Shapefile
17

17

3.1.2 Create Database Diagram
The creation of the database diagram is a crucial step in beginning the Wake County
District Overlay project. The diagram, depicted in Figure 3, includes all data that makes up the
web application, as well as all relationships between those tables. As the database powers the
entire web application, it is critical that the data and all relationships are correctly accounted for
in the design, and the database diagram ensures that. The primary keys for each table are listed
in red italics, and the foreign keys are listed in blue. Blue italics indicates the attribute is both a
foreign key and part of the primary key. For instance, in the Vote History table, the primary key
is both the Voter_Registration_Number and the Election_Date. The foreign key, linking the table
to the Voters table, is simply the Voter_Registration_Number attribute.
18

18

Figure 3 Database Diagram

Congressional
Districts (polygon)
March 2013
• District (Integer)
• Representative (Text)
• Phone
• Website (Text)
• Political_Party (Text)
State House
Districts (polygon)
March 2013
• ID (Integer)
• District (Integer)
• Representative (Text)
• Phone
• Political_Party (Text)
• Website (Text)
State Senate
Districts (polygon)
March 2013
• ID (Integer)
• District (Integer)
• Senator (Text)
• Phone
• Political_Party (Text)
• Website (Text)
Streets (line)
May 2014
• Street_ID (Integer)
• Street_Name (Text)
• Left_Zipname (Text)
• Right_Zipname (Text)
• Zip_Left (Integer)
• Zip_Right (Integer)
• Classname (Text)
• One_Way (Text)
• Class (Integer)
• State (Text)
School Board
Districts (polygon)
March 2013
• District (Integer)
• Member_Name (Text)
Precincts (polygon)
November 2013
• Precinct (Integer)
• County (Text)
FCC Media
Markets (point)
June 2012
• Callsign (Text)
• Service (Text)
• Television_Channel
(Integer)
• Status (Text)
• City (Text)
• State (Text)
• Country (Text)
• Latitude (Double)
• Longitude (Double)
• F ACID (Integer)
• Licensee (Text)
FCC Media
Markets Coverage
(polygon)
June 2012
• Call (Text)
• Prefix (Text)
• Service (Text)
• ID (Integer)
• Licensee (Text)
• Status (Text)
• State (Text)
• City (Text)
• Television_Channel
(Integer)
• Longitude (Double)
• Latitude (Double)
• Contour (Integer)
Voters (point)
June 27, 2014
• V oter_Status (Text)
• Last_Name (Text)
• First_Name (Text)
• Middle_Name (Text)
• Voter_Registration_Nu
mber (Integer)
• Address (Text)
• Address_Unit (Text)
• State (Text)
• Zipcode (Integer)
• Mail_Address (Text)
• Mail_City_State_Zip
(Text)
• Race (Text)
• Political_Party (Text)
• Gender (Text)
• Age (Integer)
• Registration_date
(Date)
• Municipality (Text)
• Precinct (Text)
• Congressional_District
(Integer)
• Superior_Court_Distri
ct (Text)
• Judicial_District
(Integer)
• NC_Senate_District
(Integer)
• NC_House_District
(Integer)
• County_Commissioner
_District (Integer)
• School_District
(Integer)
Superior Court
Districts (polygon)
November 2011
• ID (Integer)
• District (Integer)
County
Commissioner
Districts (polygon)
October 2011
• ID (Integer)
• District (Integer)
Commissioner_Name
(Text)
1..*
1..*
1..*
1..*
1..*
1..*
*..1
*..1
Polling Places
(point) March 2014
• ID (Integer)
• Precinct (Integer)
• Street_Name (Text)
• City (Text)
• Polling_Place (Text)
• Street_Number (Text)
• Full_Address (Text)
• State (Text)
1..1
1..1
Vote History
June 27, 2014
• Voter_Registration
_Number
• Election_Date
• Method_of_Voting
1..1
Vote History
Key
June 27, 2014
• Letter
• Voting_Method
1..1
19

19

3.1.3 Format Data
All of the shapefiles listed above can be added to the PostGIS database in their original
format, but the voter file spreadsheet, downloaded from the Wake County Board of Elections,
requires additional steps: geocoding the addresses and creating a separate vote history table.
First, the addresses of each voter must be geocoded in order to display on the map within the
Wake County District Overlay application. While the rest of this project utilizes open source
tools, geocoding the addresses requires the use of ArcGIS.The Wake County Streets file, listed
above in Table 1, was used to create the address locator for the ArcMap geocoder. The address
locator is dual range, allowing it to determine the correct side of the street for each address, as
shown in Figure 4.

20

20

Figure 4 Create Address Locator

The geocoding tool in ArcMap uses the address locator to determine how to visually
display the address of the voters from the voter file. The “Address,” “City,” “State,” and “Zip”
columns from the voter file input correspond to the respective street names and addresses
contained within the Wake County streets file in the address locator. With over 39,000 addresses
to geocode, that process takes several minutes for the process to complete. This file results in
39,252 (99%) matched addresses. The other 1% of addresses are not geocoded for various
reasons, including the portion of North Carolina State University students within House District
34 who are registered at “0 NCSU,” the homeless who are allowed to draw a map of where they
21

21

spend most of their time, and a few street name discrepancies between the streets file and the
voter file.
In the second phase of reformatting data, the vote history information, available in the
original voter file spreadsheet, must be divided into a separate table to be added to the database.
The table must be reformatted to include only three columns: Voter_Registration_Number,
Election_Date, and Voting_Method. To do this, a python script, written by Patty Jula, a USC
GIST student, and Bill Kronholm, a Whittier College math professor, is used to loop through
each column, row-by-row, and format the data into the previously mentioned columns (see
Appendix A). For example, the original voter file spreadsheet includes one row per voter, and a
separate column for each election, along with the relevant demographic information. In order to
reference voters by election in the database, a separate vote history table is necessary. In order to
create this reformatted spreadsheet, the above mentioned python script is utilized to move the
data into new columns. This python script can be used in future updates of the application as new
voter files are released. In the reformatted Vote_History table, there is a row for each voter for
each election they have voted in; therefore most voters have multiple rows. This is the preferred
method of storing vote history information in a database, and allows for updating the database as
elections occur. This way, new rows are added to the Vote_History every time there is an
election instead of adding a new column for each election. The Voter_Registration_Number
column serves as the foreign key between the Vote_History table and the previously created
Voters shapefile.
3.1.4 Create Database in PostGIS
After all of the data is correctly formatted, they are uploaded into the PostGIS database.
Making sure to assign the correct Spatial Reference System Identifier (SRID) (2264 for the
22

22

Wake County specific files and 4267 for the country-wide files) to all shapefiles to ensure
consistency of projection, the pgShapeloader interface is used to upload all of the shapefiles into
the database (OpenGeo Suite: pgShapeloader). pgShapeloader allows for importation of all
shapefiles into the PostGIS database through a graphical user interface instead of through
command line tools. The vote history table and the vote history key table, the only non-spatial
pieces of data, are imported directly through the PostGIS interface. Once all of the data is
uploaded, all of the relationships (referred to as constraints in PostGIS) must be set. This
includes setting all of the respective primary keys and foreign keys as determined during the
database diagram design phase. The resulting PostGIS database is shown in Figure 5.

Figure 5 PostGIS Database
23

23

3.1.5 Upload Database to GeoServer
The next step is to connect GeoServer with the PostGIS database and upload the
necessary layers. The first step in this process is to create a new workspace in GeoServer. This
communicates to GeoServer which database to connect to in PostGIS. In this case, the database
is named WakeCounty, as is shown in Figure 6.

Figure 6 Set up Workspace in GeoServer
24

24

Once the workspace is established, the next step is to upload the necessary layers. This
project uses two methods to upload the data to the PostGIS database. First, the shapefiles that are
loaded into PostGIS as tables are connected directly to GeoServer, as GeoServer is able to
recognize these layers as having spatial data. GeoServer automatically computes the bounding
boxes, latitude, and longitude for each layer and readies them for web display. Figure 7 displays
the connection between GeoServer and the WakeCounty PostGIS database, along with the
published layers.
Figure 7 GeoServer Layers

25

25

The second method of uploading layers to GeoServer is through the Structured Query
Language (SQL) view layers option. This method incorporates any changes in the database every
time the web page is loaded, forcing GeoServer to query the database before displaying the layer
in a web map. This means that as voters are added to the database, the SQL layers are able to
incorporate those new voters into the appropriate layers. Figure 8 shows the SQL query for the
layer that displays voters who voted in the 2008 general election but did not vote in the 2012
general election. This SQL layer can be reused within this instance of GeoServer and with this
PostGIS database to query voters within any given election. This only portion of the query that
needs to change is the election date. The next subchapter shows the resulting layer from this
SQL statement.

Figure 8 Example SQL View Layer

26

26

3.1.6 Build Web Application Using Leaflet
Now that all of the layers have been loaded into GeoServer, the next step is to build the
webpage, which displays the map. This project uses Leaflet, an open source JavaScript-based
library for creating interactive maps for web and mobile platforms. The first step is to link to
both the Leaflet JavaScript library and the Cascading Style Sheets (CSS) file, which informs the
browser how to display the information called in the web page. Next, the div, or section, that
holds the map is created and the map size is set to a width of 100% and a height of 600 pixels.
The first data set to add to the map is a basemap. Mapbox,
6
another open source mapping
application, allows for the creation of customized basemaps that can be added to a page powered
by Leaflet. This application uses the standard Streets basemap available from Mapbox, shown in
Figure 9.

Figure 9 Web Map Showing Mapbox Basemap

6
https://www.mapbox.com/, an open source mapping application.
27

27

Finally, all of the layers added to GeoServer in the previous subchapter are added to the
map through the JavaScript code, and are divided among their respective layer switchers. The
application comprises 32 layers that can be turned on or off by the user, including all of the
original shapefiles mentioned above, plus SQL layers allocated based on voting history, party
affiliation, Congressional District location, gender, age, and voting method.
Upon initially opening the final web application, the map loads with two layers visible:
Congressional Districts and All Voters (as mentioned in Table 1, “All Voters” in this map refers
to all Democrats and Unaffiliated voters in state House district 34, as only the Democrats and
Unaffiliated voters were geocoded through the ArcMap geocoding process). This gives a brief
overview of the county, state House district 34, and the distribution of voters in the area, as seen
in Figure 10.

Figure 10 Web Map Showing US Congressional Districts in Wake County and All
Democratic and Unaffiliated Voters in HD34 Displayed in the Wake County District
Overlay Web Application

28

28

For a more in-depth look at the political districts in the county, the State House Districts
and the State Senate Districts layers are both turned on. As is seen in Figure 11, these three
district types are not nested, meaning they do not share common boundaries. From a political
standpoint, nested districts are much easier for campaigning and voter targeting. For instance, if
one state Senate district were to comprise two state House districts, the three candidates of the
same party could coordinate their field efforts and canvass (knock on doors) for two candidates
at a time, rather than only one. This would result in the need for fewer staff and volunteers, and
would ultimately save the campaigns time and money. However, these idyllic nested districts do
not exist in North Carolina, therefore a map like the one below is incredibly useful for ensuring
campaigns are reaching the voters that are within their districts.
Figure 11 Web Map Showing US Congressional Districts, State House Districts, State
Senate Districts in Wake County and All Democratic and Unaffiliated Voters in HD34

29

29

Moving on to the next category of layers available in the Wake County District Overlay,
the user has the option to display voters based on elections in which they have or have not voted.
In the 2012 Presidential election, President Obama lost North Carolina by fewer than 100,000
votes. From a voter targeting perspective, those voters who voted in the 2008 election, but did
not vote in the 2012 election, are good targets because they have shown they will vote when
motivated. An additional face-to-face conversation with a volunteer from a campaign may be the
extra incentive they need to turn out on Election Day. And by displaying these sporadic voters on
the map, using the “YES2008_NO2012” layer, a field worker can determine where the higher
concentrations of these voters are in order to make the process as efficient as possible,
demonstrated in Figure 12. As discussed above, this is a SQL view layer, which queries the
database with the pre-loaded SQL statement before displaying.
Figure 12 Web Map Showing US Congressional Districts and Voters in HD34 who voted in
the 2008 general election, but did not vote in the 2012 general election

30

30

Another way the Wake County District Overlay application can be used is to display the
effects of redistricting on the make up of the districts. For instance, Figure 13 shows the voters
color-coded by the Congressional District in which they are located. All of these voters are
located within the same state House district (HD34), and while a majority are within
Congressional District 4, shown in purple, a small percentage of the voters are located within
Congressional District 13, shown in green. Every ten years after the census is completed, federal
and state political districts are redrawn to account for the population changes. Again, this is
where district nesting could become a huge benefit to campaigns, if the districts were drawn to
incorporate voter targeting as a parameter.
Figure 13 Web Map Showing US Congressional Districts and Voters in HD34 divided
between Congressional District 4 (purple) and Congressional District 13 (green)

The application can also show voters based on past voting methods. For instance, Figure
14 shows voters who have ever used “One-Stop Early Voting.” Early voting allows voters who
are unable to vote on Election Day to cast their vote during a set period of time beforehand. In
31

31

April of 2014, the North Carolina legislature passed a law to shorten the early voting period by
seven days, which is expected to cause long lines at the polls on Election Day (Leslie 2014).
With fewer days to vote, and longer lines at the polls, these voters may need an additional
contact from the campaign to ensure they know of the new early voting time period, and to
ensure they allow plenty of time if they plan to vote on Election Day. By displaying these voters
on a map, the campaign staff can visualize where these voters are and maximize the use of their
volunteers in the field.

Figure 14 Web Map Showing State House Districts, Precincts, and Voters in HD34 who
have ever voted using One-Stop Early Voting

Finally, the Wake County District Overlay application can be used to display the FCC
media market coverage to determine if television advertisements will reach the expected
audience. The circles represent coverage areas for each television transmitter. People located
within these circles should be able to receive those signals on their televisions. As is shown in
Figure 15, Wake County has a very high density of television markets in the area, and its
32

32

placement in the center of the state allows voters in the area to receive signals from television
stations from throughout the state. This becomes even clearer when compared to the media
market coverage in the western part of the country, as depicted in Figure 16 , where there are
areas that receive no television signals. If a future iteration of this project were implemented for
a national campaign, viewing these differences in media coverage would be crucial in the
advertising decision process.

Figure 15 Web Map Showing FCC Television Media Market Coverage in Wake County

33

33

Figure 16 Web Map Showing Nationwide FCC Television Media Market Coverage

3.2 CartoDB: Alternative to PostGIS/GeoServer/Leaflet for Web Application Design
While Leaflet allows for the creation of visually appealing interactive maps, it does not
have flexibility of functionality for the user without writing extensive server-side code; it
requires that all of the layers be pre-loaded by the administrator. While pre-loaded layers may
ultimately be preferable for this application when implemented in the field, an attempt at using
CartoDB to create an interactive web map with more user-side functionalities is necessary to
thoroughly explore the development options. CartoDB is a cloud-based mapping tool for creating
web maps that are powered by a spatial database. It allows for ease of loading and querying data,
visualization, and online production.
The first step is to load all of the data into the CartoDB database. The user interface
provides an easy drag-and-drop method for loading the files, and then each resulting table can be
edited directly through CartoDB. Once the data is optimized, the next step is to create a
visualization from a table. Up to 4 layers can be added to the visualization, and all can be queried
34

34

through the SQL interface in either the data view, which returns the resulting columns, or the
map view, which returns the resulting features.
Once all of the data is loaded into CartoDB, the final step is to build the webpage. This
version of the Wake County District Overlay differs from the original in that it includes a time
slider that displays the voting patterns of the voters over time. The html page begins with css to
define the style of the page, map, slider, and legend. Then the document calls the css stylesheets
and JavaScript libraries for both CartoDB and JQuery (required to show the time slider). The
body of the document loads the time slider and ties it to the Date column in the Vote History
table, using the following SQL query:

sublayer.setSQL("
SELECT Voters.cartodb_id, Voters.the_geom, Voters.the_geom_webmercator,
Voters.address2, Voters.arc_city, Voters.arc_state, Voters.arc_zip,
Voters.precinct, Voters.vrcongress, Voters.vrfirstnam, Voters.vrlastname,
Voters.vrgender, Voters.vrnchouse, Voters.vrncsenate, Voters.vrparty,
Voters.vrschooldi, Vote_History.date, Vote_History.voter_registration_number,
Vote_History.vtngmthd
FROM Vote_History
INNER JOIN Voters
ON Vote_History.voter_registration_number = Voters.VRN
WHERE Vote_History.date >= '" + start + "' and Vote_History.date <= '" + end +
"'");

This query joins the Vote History table and the Voters table on the Voter Registration Number
column, selects appropriate columns, and finds the voters who voted in each specific election.
When the user moves the time slider, the voters who voted in the respective elections are
displayed on the map, giving the user an overview of the voting patterns of the area, as shown in
Figure 17.
35

35

Figure 17 CartoDB Web Map Alternative to Leaflet Map

CartoDB is extremely user-friendly, both from the creation side and the user side. It is
very quick and easy to load all of the data into the database, which is then accessible from any
computer connected to the Internet, yet allows the creator to avoid setting up a server. The user
interface is very clean, with easy wizards for styling and map production. Additionally, the
CartoDB Application Programming Interface (API) requires much less JavaScript code than
Leaflet in order to create an interactive web map. However, these perks do not overcome
CartoDB’s drawbacks. With an unpaid account, a user is limited to 10 tables, and can only
display four layers on a single map. While the Leaflet Wake County District Overlay has over 30
layers available for display, the CartoDB version is limited to four: Congressional Districts, State
House Districts, Voters, and the Vote History table which does not include spatial data. For a
project that aims to show the various political districts overlaid with queried voters, this version
created using CartoDB does not meet the intended purpose.
36

36

Additionally, the JavaScript time slider does not interact very well with the data in the
CartoDB database. The slider shows voting patterns over time, with each election date showing
the voters who voted in that election. However, the time slider can only be created as a range of
dates, requiring the user to move both ends of the slider to center over a specific election date.
This does not meet the intended purpose of creating an application that aids in voter targeting,
and is not easy to use. Therefore, despite the flexibility limitations of the
PostGIS/GeoServer/Leaflet combination, that map is the better application, meets the intended
purposes set forth in Chapter 1, and is the final product put forth in this thesis project. From here,
any reference to the Wake County District Overlay application refers to the web map powered by
Leaflet.

37

37

CHAPTER 4: EVALUATION
This section details the outside evaluation of the Wake County District Overlap application by 10
professionals ranging from a GIS specialist for the state of North Carolina with over 10 years of
experience, to field staff, state data managers, senior staff, and executive level staff at AV. Each
evaluator viewed the Leaflet web map and its various functions, with several evaluators using the
application individually, and several viewing a demonstration in a group setting. Then the
evaluators answered a brief Google Survey,
7
as shown in Appendix B. The survey is divided into
three sections. The first is to gauge whether or not the application meets the intended purpose as
described in Chapter 1. The second is to demonstrate the background experience of these
evaluators to ensure that they are the intended audience or qualified reviewers. The third section
provides several open-ended questions that allow for more discussion of the thesis project.
4.1 Survey Results on Evaluator Experience
As this project is geared towards political staff who may or may not have formal GIS
training, it is important to assess whether the respondents are indeed part of the target audience
for this web-based application. Again, the survey ask three questions about experience on a scale
of 1 to 5, with 1 being 0 years and 5 being 10+ years. The first question asks, “How much GIS
experience do you have?” with only one respondent indicating 10+ years of experience. Figure
18 shows the distribution of GIS experience among respondents, with the majority falling at the
lower end of the spectrum. As is indicated by the responses to this question, the reviewers who
work for AV are on the lower end of the scale in terms of GIS experience. The answer of “5”, as
shown below, is the response from the GIS Specialist for the state of North Carolina.

7
https://docs.google.com/forms/d/1d6iTTlwsBDbZJw6cxMQL4ojv1ak47C5mkAbd0f4ZkTA/edit?hl=en
38

38

Figure 18 Graph of Survey Question #4 Answers

Conversely, the next two questions deal with campaign experience and have a much
different distribution from the previous question. The next question asks, “How much political
experience do you have?” with 40% of respondents indicating 10+ years of experience. This
shows that the intended audience of professional political campaign staff has been met by the
sample of respondents, with all evaluators having at least some political experience.
Additionally, this small sample includes people in various stages of their careers, ranging from
those with over 10 years of experience at the executive level to those who are just beginning
their political careers. This is important, as it critical to include opinions from seasoned
professionals who have a vast amount of experience as well as from those early in their careers
who may have new ideas to add to the field. See Figure 19 for a graph of the answers.
39

39

Figure 19 Graph of Survey Question #5 Answers

The third question asks, “How much field experience do you have?” This differs from the
previous question in that it is possible to have political experience without field experience. For
example, political fundraisers may indicate that they have political experience, but do not have
field experience. On a campaign, the Field Department is the outreach for the campaign; field
staffers call voters, knock on doors, and organize volunteers to help persuade voters and turn
them out on Election Day. This application is geared more towards political staff who work in
field. As is indicated in Figure 20, this group of respondents has a wide range of field experience,
ranging from the National Field Director and National Political Director at AV to the state Data
Managers and even the Development Associate.
40

40

Figure 20 Graph of Survey Question #6 Answers

4.2 Survey Results on Gauging Whether the Wake County District Overlay Meets the
Intended Purpose
The survey asked several questions of respondents to gauge whether the application
meets the intended purpose set forth in Chapter 1. The first question asks, “How useful would
this web map be for voters targeting?” on a scale of 1 to 5, with 1 being “Not Useful” and 5
being “Useful.” All evaluators responded with either a 4 or 5, demonstrating that the application
is a worthwhile tool. Seventy percent of evaluators answered “4”, indicating that there is always
room for improvement, but that the application as it exists would still prove to be useful. The
potential improvements are outlined in Chapter 5. Figure 21 shows a graph of the responses to
this question.
41

41

Figure 21 Graph of Survey Question #1 Answers

The next question, again on a scale of 1 to 5, with 1 being “Not At All” and 5 being
“Very Well” asks respondents, “How well does this map help you visualize where the low-
propensity voters are?” 80% of evaluators answered 5, Very Well, indicating that the application
also meets the purpose of visualizing the low-propensity voters. One evaluator responded with a
“2” on the ability of this thesis project to aid in visualizing where the low-propensity voters are,
yet also rated the previous question of usefulness of the web map as a “4”. As discussed in
Chapter 5, this reviewer has several suggestions for building on the capabilities of this project.
The answers to this survey question are displayed on a graph in Figure 22.
42

42

Figure 22 Graph of Survey Question #2 Answers

Finally, the third question in this intended purpose section asks, “Is this web map easy to
use?” on a scale of 1 to 5, with 1 being “Not Easy” and 5 being “Very Easy.” All evaluators
responded with 3 or higher, which is shown in Figure 23. As indicated in the previous section,
90% of the respondents indicated a level “3” or lower in terms of GIS experience, while
indicating in this question that the web map is easy to use. According to the evaluators, the goal
of creating a web application that is easy to use, even for the untrained, is met.
Figure 23 Graph of Survey Question #3 Answers

43

43

4.3 Survey Results from Open-Ended Questions
The final section of the survey included three open-ended questions to allow for more in-
depth answers by the evaluators. The first question in this section asks, “Do you use other tools
that already provide this functionality?” Four respondents who are employed by AV indicated
that the current tools utilized by the organization provide limited mapping functionality, some of
which may overlap with this application, but are not as comprehensive. Additionally, Josh
Uretsky, Data Manager for AV-Pennsylvania, stated, “I have used qGIS
8
[an open source
geographic information system] or other mapping software to visualize targeting, but using
something like this would be significantly easier and quicker” (Uretsky 2014). This is in
accordance with the stated intent from Chapter 1 that the application “allows a non-GIS
professional to quickly and easily display the various spatial data.”
The second open-ended question asks, “Are there any changes you would make?
Additions? Deletions?” Most respondents agreed that they would like to build upon this
foundation, creating an even more powerful voter-targeting application in subsequent iterations.
Several of the suggestions brought forth in the survey, including export functions, turf cutters,
connecting to the existing database, and adding more commercial data are discussed in Chapter
5.
The final open-ended question asks for additional comments that the survey may not have
asked. Josiette White, National Field Director for America Votes, indicated that the application
has clear advantages over other applications currently utilized, stating, “One advantage over
other tools: I do not need to take multiple steps to get to the visualization. Most other systems
require multiple other steps before you can visualize the voters” (White 2014). As stated in
Chapter 1, making the process quick and efficient is key to winning campaigns. In terms of

8
www.qgis.org/, an open source Geographic Information System
44

44

sharing the maps created by the application with partner organizations, Toby Miceli-
Gwiazdowski, AV-Wisconsin Data Manager, asserts that the Wake County District Overlay
“would be perfect for a presentation at a table meeting to be able to visualize and answer
questions in real-time” (Miceli-Gwiazdowski 2014). Finally, Sara Schreiber, Managing Director
of America Votes states, “This seems like a great product that could be refined into something
that our field department could use all the time” (Schreiber 2014).
The evaluation design has several strengths and limitations. All of the evaluators, with
the exception of the GIS programmer, fell within the intended audience of current political staff
who may choose to use this application as a tool for targeting voters. The evaluators were able to
give feedback about the application in relation to their current jobs and how the application could
be integrated into their work plans. Additionally, as the staffers have various levels of GIS
experience, the application proved easy to use regardless of this level of experience. However,
the sample size for the evaluation was fairly small with only 10 respondents, 9 of whom work for
the same organization. Future testing of the application will require a larger sample size of
political staffers. Additionally, if this application is to be used outside of America Votes, future
evaluations will need to include staffers who work for other organizations to ensure
compatibility across the board.

45

45

CHAPTER 5: DISCUSSION AND CONCLUSION
This section discusses the potential impact of the Wake County District Overlay application on
voter targeting in political campaigns. Additionally, this section discusses the shortcomings of
the application and lays the groundwork for future enhancements.
5.1 Results of the Wake County District Overlay Web Application
Overall, the Wake County District Overlay was very well received, with everybody in
agreement that the application met its intended purpose of aiding in targeting of low-propensity
voters. Many would like to see this project, or future iterations of the project, implemented in the
field. The application gives an excellent overview of the voters within state House district 34 in
terms of voting habits, demographic trends, and even impact of redistricting and voting rights
laws on the electorate. By displaying the sporadic voters on a map, campaign staffers are able to
visualize where those voters are located and build canvassing plans off of those visualizations.
In the database currently utilized, campaign staff can export a list of those sporadic voters
into a spreadsheet. However, viewing a list of 2000 names, the approximate number of Democrat
and Unaffiliated voters in state House district 34 who voted in the 2008 general election but not
the 2012 general election, does not have the same impact as seeing those voters visualized on a
map. Seeing these voters displayed on a map in the Wake County District Overlay can be used to
visualize the walk routes and ensure the canvassers use the most efficient routes possible to reach
all voters. The same is true of the demographic trends and the policy impacts. Seeing these voters
on a map – envisioning exactly where they live, what their neighborhoods look like, and how
they are distributed through the district – gives a better sense of the distribution of voters than a
spreadsheet.
46

46

Additionally, the design of the Wake County District Overlay application allows for easy
updating of both the database and the user interface. By using the aforementioned SQL view
layers to create the various overlays, the application is ready for database updates. Before
displaying each of these layers, GeoServer queries the PostGIS database with the pre-loaded
SQL statement, which ensures that any updates to the database are always included in the
refreshed view of the web page. Wake County publishes updated voter files every week, so the
database powering the Wake County District Overlay could potentially have updates of newly
registered voters added on a weekly basis. This guarantees that campaigns are working off of the
most up-to-date information and are able to reach all of the voters they need to before Election
Day. Furthermore, as each election occurs, new vote history information will need to be added to
the vote history table. The python script detailed in Chapter 3 can be used again to format the
data from the format given on the Wake County Board of Election website into the format
required by the database. This updated information can then be added directly into the PostGIS
database. In addition to the ease of database updates, the user interface can be easily updated as
well. If a field staffer were to request a layer not already available in the Wake County District
Overlay, the application administrator could very easily add those layers at the staffer’s request.
5.2 Future Improvements to the Wake County District Overlay
As with any development project, there are always ways in which the project can be
amended and improved. To extend this application statewide, the PostGIS database and
GeoServer would need to be installed on a server, rather than running locally on a laptop. This
would allow a campaign to target voters throughout North Carolina, rather than in a specific
House District. The database design would also need to incorporate data for all 100 counties in
47

47

the state, rather than just for Wake County. The design could then be replicated to other states as
well.
This application could also be improved by the addition of new functionalities. The first
step toward adding on to this project will be to include data export capabilities, as noted by Ed
Shipman in the survey detailed in Chapter 4, a GIS Programmer who reviewed the Wake County
District Overlay application (Shipman 2014). Ideally, the user could export both the map and the
voter data that corresponds to the extent visible on the screen. This would allow the user to easily
put together packets of information about all of the voters they had queried, and would include
vote history, address, and demographic data. Adding in this new capability requires designing
new functionalities for the web application. It would require implementing a way for the user to
export the data which exists in the database. If this were added, and if the web application were
hosted on the Internet and publicly available, extra security measures would also need to be
included to ensure the integrity of the data and to prevent unauthorized users from gaining access
to the database.
The next logical step, after export capabilities, is to include a turf cutter, an addition
suggested by Jack Nguyen, Deputy Data and Targeting Director at America Votes, in response to
the survey (Nguyen 2014). In campaign field departments, “cutting turf” refers to creating walk
maps of about 40 houses to give to staffers or volunteers to canvass. These usually include a map
and the voter information included above, plus any previous contacts. However, the current
method of cutting turf involves a lot of manual intervention on the part of the field staffer. If
cutting turf were completely automated, the process would be much faster and more efficient,
saving time for the staffer. This potential improvement would require extensive work to add new
capabilities for the application. The application would need to be able to group the geocoded
48

48

voters into groups of about 40 houses in order to create the walk maps. Additionally, the houses
would need to be easily canvassed by staff or volunteers, typically by walking door-to-door. A
walkability assessment may also be necessary in order to complete this new step, ensuring that
the canvassers would be able to easily access all of the houses on a given walk map.
Another important step in optimizing the Wake County District Overlay is to connect to
the existing database already utilized by campaigns. This step, recommended in the survey by
both Evan Kost and Josh Uretsky of America Votes, would eliminate the need for PostGIS and
would improve map rendering times, as the existing database for the organization is housed on a
powerful server instead of the PostGIS instance running on a laptop (Kost 2014; Uretsky 2014).
This would also give the application access to all voters in the database, not just the voters in a
specific House district, and would in turn provide the capability of expanding this project
statewide. The implementation of this suggestion would require the same steps as the workflow
detailed in this paper, yet would not require the use of PostGIS.
Finally, several evaluators mentioned the possibility of adding in additional data to the
application. Paula Hodges, AV-New Hampshire State Director, in response to the survey
suggested adding in more commercial data, such as whether voters own or rent their homes, or
even home values, as an added parameter for querying the voters (Hodges 2014). As she
mentions, people who own their homes, as opposed to renting, often have different voting
opinions and tendencies. Additionally, Josh Uretsky would like to see Census data added in as an
additional indicator of demographic trends throughout the area (Uretsky 2014). Given the current
database design, adding these data would be easy to complete. It would require the same steps as
the workflow detailed in the previous chapters, and would require adding new tables to the
49

49

PostGIS database. These new tables could be linked to GeoServer in the same way as the current
tables and then displayed within the Leaflet map.
The methodology behind the creation of this application not only creates a valuable
voter-targeting tool, but also provides a framework for easy replication and extension. By
following the workflow detailed in Chapter 3, the application can be duplicated to fit any State
House District, as long as the necessary data is available. Because the PostGIS database,
GeoServer, and Leaflet are open source, any potential duplicator has access to all necessary
components. While the geocoder is a tool within ArcGIS, proprietary GIS software, other
geocoding options exist, in the event that the replicator does not use Esri products. Furthermore,
extending the application statewide would simply require expanding the database design to
encompass all voters and districts within the state. This would allow for use of the application on
multiple campaigns at all levels of the ballot.

50

50

REFERENCES
Altom, T. 2012. Political Parties Use Data Mining to Value You. The Indianapolis Business
Journal.
ArcGIS. Geographic Information System software developed by Esri.
http://www.esri.com/software/arcgis
Blaemire, B. 2013. Evolution of the Voter File. Campaigns & Elections 318: 21-23.
CartoDB. “CartoDB Documentation.” Last accessed 24 August 2014. https://cartodb.com/docs.
Cho, J. 2011. The Geography of Political Communication: Effects of Regional Variations in
Campaign Advertising on Citizen Communication. Human Communication Research 37:
434-462.
Cho, W, and J. Gimpel. 2009. Rough Terrain: Spatial Variation in Campaign Contributing and
Volunteerism. American Journal of Political Science 54: 74-89.
Cho, W, and J. Gimpel. 2012. Geographic Information Systems and the Spatial Dimensions of
American Politics. Annual Review of Political Science 15: 443-460.
Darmofal, D. 2006. The Political Geography of Macro-Level Turnout in American Political
Development. Political Geography 25: 123-150.
Delany, C. 2011 Online Advertising: A Targeting-Rich Environment. Campaigns & Elections
307: 15-16.
Delany, C. 2012. Voter Files Digital Ad Targeting: Reality vs. Hype. Campaigns & Elections
312: 20-21
Esri. “Esri Shapefile Technical Description.” An Esri White Paper. July 1998.
Federal Communications Commission GIS. Last accessed June 2014.
http://wireless.fcc.gov/geographic/index.htm?job=licensing_database_extracts
51

51

GeoServer. “GeoServer Documentation.” Last accessed 24 August 2014.
http://docs.GeoServer.org/stable/en/user/.
Gill, C. Personal communication to author. 2 April 2014.
Hodges, P. Google Survey to author. 22 August 2014.
Jessee, S. A. 2010. Partisan Bias, Political Information, and Spatial Voting in the 2008
Presidential Election. The Journal of Politics 72.2: 327-340.
Kacor, B. 2012. Voter Registration Drives Adopting New Methods. The Louisiana Weekly 24
September.
Karp, J. A., S. A. Banducci, and S. Bowler. 2008. Getting Out the Vote: Party Mobilization in a
Comparative Perspective. British Journal of Political Science 38.1: 91-112.
Keschmann, M. 2013. Reaching the Citizens: Door-to-Door Campaigning. European View 12.1:
95-101.
Kost, E. Google Survey to author. 22 August 2014.
Leaflet. “Leaflet API.” Last accessed 24 August 2014. http://leafletjs.com/reference.html.
Leslie, Laura. 2014. Shorter Early Voting Period Could Bring Longer Lines for NC Primary.
WRAL 4 April.
Lilienthal, S. 2005. Targeting the Individual Voter. Campaigns and Elections 26: 39-40.
Luebke, T. Email message to author, 28 March 2014.
McKee, S. C. and J. M. Teigan. 2009. Probing the Reds and Blues: Sectionalism and Voter
Location in the 2000 and 2004 U.S. Presidential Elections. Political Geography 28.8:
484-495.
Miceli_Gwiazdowski, T. Google Survey to author. 22 August 2014.
Nguyen, J. Google Survey to author. 22 August 2014.
52

52

OpenGeo Suite: pgShapeloader. “Loading Data into PostGIS with pgShapeloader.” Last
accessed 24 August 2014
http://suite.opengeo.org/4.1/dataadmin/pgGettingStarted/pgshapeloader.html.
OpenGeo Suite. http://boundlessgeo.com/.
Popper, B. 2013. As Obama Heads Back to Office, A Battle Rages Over the Tech that Got Him
Reelected. The Verge. Last accessed 5 September 2014.
http://www.theverge.com/2013/1/22/3902746/obama-heads-back-office-battle-rages-
over-tech-that-got-him-reelected
The PostGIS Development Group. PostGIST 2.1.4dev Manual. Last accessed 24 August 2014.
http://postgis.net/docs/manual-2.1/.
Python. A computer programming language. https://www.python.org/.
Rigamer, G. 1998. Targeting and Predicting Voter Turnout. Campaigns and Elections 19: 42-45.
Schlozman, K. L., S. Verba, and H. E. Brady. 2010. Weapon of the Strong? Participatory
Inequality and the Internet. Perspectives on Politics 8.2: 487-509.
Schreiber, S. Google Survey to author. 22 August 2014.
Sheehan, M. 2012. On the Campaign Trail: Matt Sheehan Examines the Phenomenal Rise in
Popularity of Mobile Devices; the Impact on Political Organization and Campaigning,
and the Game-changing Role for GIS. Business Insights: Essentials.
Shipman, E. Google Survey to author. 22 August 2014.
Sui, D. and Peter J. Hugill. 2002. A GIS-based Spatial Analysis on Neighborhood Effects and
Voter Turn-Out: A Case Study in College Station, Texas. Political Geography 21: 159-
173.
Uretsky, J. Google Survey to author. 22 August 2014.
53

53

Wake County Map Services. Last accessed June 2014.
http://www.wakegov.com/gis/services/Pages/data.aspxWake County Voter Data. Last
accessed 27 June 2014. http://www.wakegov.com/elections/data/pages/data.aspx
WakeGOV People and Places. Last accessed 7 September 2014.
http://www.wakegov.com/planning/peopleandplaces/Pages/default.aspx.
White, J. Google Survey to author. 22 August 2014.
White, J. S. and B. Nuckels. 2013. Targeting Cable with Precision. Campaigns and Elections
316: 29-31.
Williams, J. 2014. Four Tactics Your Campaign Should Take Advantage of in 2014. Campaigns
& Elections 322: 16-17.

54

54

APPENDIX A: PYTHON SCRIPT FOR REARRANGING DATA IN EXCEL
Authors: Patty Jula and Bill Kronholm

55

55

APPENDIX B: GOOGLE EVALUATION SURVEY

56

56

Abstract (if available)

Abstract Political campaigns are inherently geography-driven. The creation of a tool that allows a non-GIS (geographic information systems) professional to quickly and easily display the various spatial data concerning voters and their locations is key to winning elections. A political non-profit organization headquartered in Washington, D.C., has expressed an interest in a web-based mapping application that aids in voter targeting, especially the targeting of registered voters with irregular voting habits in order to persuade them to vote on Election Day. This thesis presents a web-based application, which provides campaigns and organizations with fast access to the knowledge they need to manage field operations. This project relies on open source software, including Leaflet, PostGIS, and GeoServer. The project focuses on Wake County in North Carolina, with the expectation of expanding the web application to the state level in the future. The data needed for the project are readily available and include publicly available voter files of all registered voters, and shapefiles of Wake County media markets, precincts, polling places, state House and Senate districts, Congressional districts, School Board districts, County Commissioner districts, and Superior Court districts. The evaluation shows that campaign staffers can use the web application to efficiently and effectively visualize relevant combinations of the above data and share this knowledge with their colleagues.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Guiding business oriented volunteered geographic information through geotigger services: a case study of CrossFit affiliates

PDF

The geography of voter power in the U.S. electoral college from 1900-2012

PDF

Semi-automated visualization of spatial information in unstructured text

PDF

Finding your best-fit neighborhood: a Web GIS application for online residential property searches for Anchorage, Alaska

PDF

Campaign financing for the U.S. House of Representatives: an interactive Web map

PDF

Visualizing sporting event temporary flight restrictions

PDF

A user study of GIS infused genealogy with dynamic thematic representation and spatiotemporal control

PDF

Comparative 3D geographic web server development: visualizing point clouds in the web

PDF

Investigating electoral college reform: geography's impact on elections, and how maps influence our perception of election outcomes

PDF

Development of a Web GIS for urban sustainability indicators of Oakland, California

PDF

The role of GIS in asset management: County of Kauai Department of Parks and Recreation a need for an asset management program

PDF

Generating bicyclist counts using volunteered and professional geographic information through a mobile application

PDF

Visualizing email response data to improve marketing campaigns

PDF

San Diego Wildfire Hazards Information Center mashup

PDF

Mapping uniformity of park access using cadastral data within Network Analyst in Wake County, NC

PDF

A fire insurance map geocoder for pre-earthquake San Francisco

PDF

GIS data curation and Web map application for La Brea Tar Pits fossil occurrences in Los Angeles, California

PDF

Mapping native plants: a mobile GIS application for sharing indigenous knowledge in Southern California

PDF

Geospatial analysis of unintended casualties during combat training: Fort Drum, New York

PDF

Analyzing earthquake casualty risk at census block level: a case study in the Lexington Central Business District, Kentucky

Asset Metadata

Creator Bunn, Haynes Hoyle (author)

Core Title Wake County District Overlay: an online electoral data visualization application

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 09/12/2014

Defense Date 09/03/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Elections,GeoServer,GIS,Leaflet,OAI-PMH Harvest,Political campaigns,Politics,PostGIS

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Swift, Jennifer N. (committee chair), Chiang, Yao-Yi (committee member), Warshawsky, Daniel N. (committee member)

Creator Email haynesbunn@gmail.com,hbunn@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-475370

Unique identifier UC11286594

Identifier etd-BunnHaynes-2927.pdf (filename),usctheses-c3-475370 (legacy record id)

Legacy Identifier etd-BunnHaynes-2927.pdf

Dmrecord 475370

Document Type Thesis

Format application/pdf (imt)

Rights Bunn, Haynes Hoyle

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA