Part III: Elections, Voting, and the Electorate
The Map That Doesn't Match
Jake Rourke had run county-level campaigns before, but the Whitfield Senate campaign was his first statewide race, and the first thing that surprised him was the map. The Garza-Whitfield battleground covered a Sun Belt state with a fast-growing metropolitan core, a shrinking but still numerous rural base, and a suburban ring that had been trending away from Republicans for three cycles. Looking at a county map colored by 2020 presidential margins, the state seemed deeply red. Looking at a population-weighted map, it looked like a genuine toss-up. Looking at a model of the likely electorate — who would actually show up on Election Day — it looked like something different still.
Three maps, three different stories, three different strategic implications. The question of which map is correct is not primarily a technical question. It is a question about what elections actually measure, which voters actually participate, and why people make the choices they do. These are the questions that political scientists have been investigating for more than seventy years, and Part III synthesizes the most important findings from that literature for an analytics audience.
The six chapters of Part III constitute the theoretical heart of the book. While Parts II and IV are primarily methodological — teaching you how to measure and predict — Part III is primarily behavioral: it asks why people vote the way they do and what that means for anyone trying to understand, model, or influence an election. No forecasting model in Part IV will make sense without the behavioral foundations laid here.
What This Part Covers
Chapter 11: The American Voter and Beyond revisits the foundational studies of electoral behavior, beginning with the landmark 1960 work by Campbell, Converse, Miller, and Stokes, and tracing the evolution of electoral behavior research through retrospective voting models, economic voting theory, and cross-national comparative work. The chapter is not a literature review for its own sake — it is an excavation of the assumptions embedded in every voter model and targeting algorithm that campaign analytics teams deploy today.
Chapter 12: Partisanship, Polarization, and Sorting examines the most consequential structural shift in American electoral politics over the past four decades: the alignment of partisan identity with racial, educational, geographic, and cultural identities in ways that have made party identification both a stronger predictor of vote choice and a more emotionally charged aspect of personal identity. For the Garza campaign's targeting team, understanding polarization is not an academic exercise — it directly determines how many genuinely persuadable voters exist in the district and where they live.
Chapter 13: Demographics and the Electorate provides the empirical baseline that every political analyst needs: how race, gender, age, education, income, religion, and geography are distributed in the electorate and how those distributions are changing. This chapter also introduces the concept of the "coalition" — the combination of demographic groups that a candidate needs to assemble to win — and explains why coalition arithmetic is more complicated than it appears in punditry.
Chapter 14: Turnout takes on what is arguably the central variable in American electoral politics. The American electorate is not a fixed entity — it is a moving target defined by who participates. Turnout in presidential elections exceeds 60 percent; in many midterms and local elections, it falls below 30. The gap between a high-turnout and low-turnout electorate can change which party wins an election without a single voter changing their preference. This chapter covers the individual, social, and structural determinants of turnout, and explains why turnout modeling is both the most valuable and the most treacherous analytical task in campaign work.
Chapter 15: Campaign Effects asks a question with a humbling answer: Do campaigns actually change election outcomes? The honest answer is: sometimes, modestly, at the margins — and those margins matter enormously in close elections. The chapter reviews the experimental and quasi-experimental literature on canvassing, phone banking, direct mail, and digital advertising, and distinguishes between persuasion effects and mobilization effects with clear implications for resource allocation. Nadia Osei's targeting model for the Garza campaign is built around exactly this distinction.
Chapter 16: Visualizing the Electorate is the part's Python chapter. Working with census data, historical election returns, and a simulated voter file, you will build a series of visualizations — choropleth maps, demographic pyramids, turnout heat maps, and shift scatter plots — that reveal the structure of a competitive electorate. The chapter introduces geopandas, matplotlib, and plotly, and by the end you will have produced a visual briefing document of the kind that campaign analytics teams prepare for senior staff at the start of a race.
Why This Sequence
Part III moves from the individual (Chapter 11) to the structural (Chapter 12) to the demographic (Chapter 13) to the behavioral (Chapter 14) to the strategic (Chapter 15) and finally to the visual-analytical (Chapter 16). This sequence mirrors the way an analyst actually approaches a race: beginning with an understanding of individual decision-making, then situating those decisions in structural context, then mapping demographic distributions, then projecting likely participation, then assessing what interventions might move outcomes, and finally producing the visualizations that make all of this legible to non-analysts.
The part also builds toward a key conceptual distinction that will be essential in Part IV: the difference between the registered voter universe, the likely voter universe, and the actual Election Day electorate. A forecasting model that does not account for this distinction is not modeling the election — it is modeling something else entirely.
Recurring Themes at Work
Prediction vs. Explanation is the dominant tension of Part III. Behavioral research on voting is primarily explanatory — it tells us why people vote as they do. Campaign analytics is primarily predictive — it tries to identify who will vote and how. These are different enterprises, and the translation between them involves assumptions that are rarely made explicit. Chapter 11 begins the interrogation of those assumptions, and it runs through every subsequent chapter.
Who Gets Counted, Who Gets Heard reaches its most politically charged expression in Chapter 14. Turnout is not a natural phenomenon. It is shaped by registration laws, polling place locations, voter ID requirements, early voting availability, and targeted mobilization — all of which have differential effects across race, class, and geography. The gap between the eligible electorate and the actual electorate is a political fact, not merely a methodological complication, and Part III treats it as such.
The Gap Between the Map and the Territory animates Chapter 13 and Chapter 16 in particular. Demographic categories in political analysis — "Latino voters," "suburban women," "non-college whites" — are tools of analysis, not descriptions of homogeneous blocs. Learning to use those categories rigorously while remaining aware of their limitations is a core analytical discipline.
An Invitation
Jake Rourke's three maps are a parable for what Part III is about. Elections are not simply the aggregation of fixed preferences. They are the product of who decides to participate, which structural forces shape that participation, how candidates and campaigns intervene, and what stories — analytical and journalistic — get told about the process. Understanding all of that requires the behavioral science of Part III. The models of Part IV will be built on this foundation. It is worth building it carefully.
Read these six chapters as a conversation between social science and political practice — each discipline making the other more honest. Then open your Python environment, load your first demographic dataset, and start mapping the electorate you will spend the rest of the book trying to understand.