Chapter 13: Demographics and the Electorate

Opening: The Numbers That Don't Lie — and the Conclusions That Do

In 2002, John Judis and Ruy Teixeira published The Emerging Democratic Majority, a book that seemed to offer Democrats a mathematical promise: demographic trends — the growth of minority populations, the expansion of college-educated professionals, the urbanization of the country — were inexorably building a durable Democratic electoral coalition. Republicans, they implied, were on the wrong side of history. Democrats just had to wait.

That argument looked prescient in 2008, when Barack Obama assembled a coalition that included record turnout among Black voters, strong performance with Latinos and young voters, and significant gains among college-educated whites. It looked shaky by 2010, when Democrats suffered massive losses in the midterms. It looked completely wrong in 2016, when Donald Trump assembled a coalition of non-college white voters substantial enough to win the Electoral College despite losing the popular vote. And then it looked half-right and half-wrong in 2020, when Biden reconstructed much of the Obama coalition while Trump made modest but real gains with Black and Latino voters.

What went wrong with the demographic destiny argument? Not the demographic facts — those were substantially correct. What went wrong was the interpretation: the assumption that demographic groups vote in permanent, fixed ways that can be extrapolated forward mechanically. Demography is not destiny. It is one force among several, and the relationship between demographic change and political change is mediated by mobilization, messaging, economic conditions, and — most importantly — the strategic choices of political parties and candidates.

This chapter examines the actual demographic composition of the American electorate, how it is changing, and what those changes mean for political analysis. We pay special attention to Theme 2 — Who Gets Counted, Who Gets Heard — because demographic analysis is never politically neutral. The decision about which groups to count, how to categorize them, and what to infer from their behavior has real consequences for whether those groups are treated as electoral targets worthy of investment or as afterthoughts.

Adaeze Nwosu and Sam Harding at OpenDemocracy Analytics have built demographic analysis tools specifically to counter the most misleading tendencies in how journalists and campaigns think about demographic groups. We'll use their framework throughout this chapter, alongside Nadia Osei's on-the-ground analysis of the Garza-Whitfield state.

13.1 What We Mean by "Electorate Composition"

The first thing to understand about demographic analysis of electorates is that "the electorate" is not a fixed thing. It has multiple layers:

The citizen voting-age population (CVAP): All U.S. citizens over 18, regardless of registration or voting history. This is the denominator for eligibility.

The registered voter population: Those who have registered to vote. This varies by state and registration rules, and is shaped by both individual choices and structural barriers.

The likely voter population: Those who are actually expected to turn out in a given election. This is what campaigns and pollsters care most about, and it differs substantially from the registered voter population.

The actual voter population: Verified by turnout records after the election. This is what exit polls and post-election analyses measure.

These four populations can look dramatically different demographically. The citizen voting-age population in many Sun Belt states is substantially more Hispanic than the likely voter population, because Hispanic citizens are younger (and young people turn out at lower rates), include more recent naturalizations, and face higher structural barriers to participation in some states. The "demographic change" that demographers observe in the population is translated into "electoral change" only through the filter of registration, mobilization, and turnout.

This is the first way that demographic destiny thinking goes wrong: it reads population trends directly into electoral trends without accounting for the translation mechanisms.

💡 Intuition: The Three Filters

Think of demographic change reaching the electorate through three filters: naturalization (for immigrants), registration, and turnout. Each filter reduces the proportion of the eligible population that actually participates. And each filter is not random — it systematically selects against younger voters, lower-income voters, and voters with less education. This means that the actual voting electorate is consistently older, wealthier, and more educated than the citizen adult population. Understanding these filters is essential for translating demographic trends into electoral forecasts.

13.2 The Education Realignment

Perhaps the most consequential demographic shift in American electoral politics over the past thirty years is the education realignment: the movement of college-educated white voters toward Democrats and non-college white voters toward Republicans. This has restructured partisan coalitions more dramatically than any other single shift and is central to understanding the contemporary political landscape.

A Brief History of the Education Divide

In the mid-20th century, education was associated with higher socioeconomic status, and higher socioeconomic status was associated with Republican voting. Business owners, professionals, and managers — who were disproportionately college-educated — leaned Republican. Working-class Americans — who were disproportionately non-college — leaned Democratic, partly through union membership and class identity.

This pattern began to crack in the 1990s. College-educated white voters started showing greater responsiveness to social and cultural issues — reproductive rights, LGBTQ+ rights, environmental regulation — on which Democrats increasingly held liberal positions. As the parties sorted ideologically and the Republican Party became more explicitly socially conservative, college-educated white voters — particularly college-educated white women — began drifting Democratic.

The 2016 election marked a threshold: for the first time in the modern polling era, non-college whites gave a Republican presidential candidate a larger margin than college whites. Trump won non-college whites by about 39 points while winning college whites by only 4 points. Biden in 2020 actually won college whites by a small margin — also a historical first.

What's Driving the Realignment?

The education realignment is not primarily about income — it is not a simple return to class voting in a new form. College-educated and non-college voters at similar income levels show different political alignments. The education divide seems to be driven by several distinct mechanisms:

Credential-based cultural identity: College graduates have been exposed to more multicultural environments, liberal arts curricula emphasizing diversity, and professional networks that skew cosmopolitan. These experiences shape values and cultural identity in ways that correlate with Democratic politics.

Exposure to institutional change: College graduates are more likely to work in sectors (tech, healthcare, education, finance) that have undergone significant demographic change and where diversity and inclusion are explicit institutional values. Non-college workers in manufacturing, construction, and resource extraction have experienced different institutional cultures.

Economic anxiety and attribution: Non-college white voters have experienced stagnant wages, deindustrialization, and community decline over the same period that college-educated workers have experienced wage growth. They have been more receptive to narratives attributing these economic outcomes to immigration, trade, and elite neglect — narratives that the Republican Party has increasingly adopted.

Social identity and perceived cultural threat: For some non-college white voters, the realignment is partly about a sense that mainstream institutions — universities, media, corporations — no longer respect their values and experiences. Republican appeals to "real Americans" and attacks on "elites" resonate with this identity.

📊 Real-World Application: Education in the Garza-Whitfield State

The Sun Belt state's educational composition creates specific dynamics for both campaigns. The major metro area contains a large and growing population of college-educated workers in tech, healthcare, and professional services — this population has been moving Democratic and constitutes a key component of Garza's base. The suburban ring includes a mix: established college-educated families who lean Republican on economic issues but have been drifting toward Democrats, and newer arrivals who skew Democratic.

The state's rural areas and mid-size cities have large non-college populations, including both white and Latino non-college voters. Nadia's data shows that non-college Latinos in the state are more persuadable than non-college whites — they have not fully sorted into the Republican coalition — but the trend is in Whitfield's direction. Jake Rourke knows this: his campaign is specifically targeting non-college Latino voters with economic messaging that emphasizes jobs and small business ownership over immigration.

The Limits of Education as a Category

Education is a powerful demographic predictor of vote choice, but it has real limitations as an analytical category:

Four-year degree as threshold: The standard "college-educated vs. non-college" distinction treats anyone without a four-year degree as equivalent — from someone who dropped out after one semester to someone with an associate's degree and 20 years of technical training. This bluntness obscures real variation.

Race and education interact: The education realignment is most pronounced among white voters. For Black voters, education has not produced a dramatic shift toward Republicans — Black college graduates remain strongly Democratic, though they may hold different issue priorities than non-college Black voters. For Latino voters, education is associated with higher income and more complex political patterns. Analyzing education as if it had the same meaning across racial groups is an error.

The credential vs. knowledge distinction: What is it about having a college degree that correlates with political behavior? Is it what you learn? Who you meet? The credential itself? The economic premium? These different mechanisms have different implications for how stable the realignment will be.

13.3 Race and Partisan Coalitions: Stability and Change

The racial composition of partisan coalitions is the most politically charged demographic topic in American politics, partly because the stakes are so high and partly because public discourse about it is often distorted in both directions — by those who overstate minority monolithism and by those who overstate minority shifts toward Republicans.

Black Voters: Stable but Not Monolithic

Black voters have supported Democratic presidential candidates at rates of 85-95% in every election since 1964, making them the most reliably Democratic demographic group in the country. This stability reflects a combination of factors: the historical association of the Democratic Party with civil rights, the continuing relevance of race as a political concern, strong community institutions (particularly Black churches) that reinforce partisan identity, and an ongoing evaluation of which party better addresses the concerns of Black communities.

The stability of Black partisan alignment does not mean Black voters are monolithic in other ways. Black voters have diverse views on economic issues, education policy, public safety, and cultural questions. The assumption that Black voters are uniformly progressive is empirically incorrect and analytically lazy — it leads campaigns to take Black voters for granted and fail to address the specific concerns of Black communities.

Recent elections have shown some variation in Black turnout and Democratic margins — Trump's Black vote share increased modestly from 2016 to 2020, particularly among Black men — but these shifts have been smaller than some commentators claimed. Careful analysis of exit poll data and validated voter records shows that large shifts in Black partisan alignment have generally not occurred and that claims of dramatic Trump gains with Black voters in 2020 were substantially overstated.

⚠️ Common Pitfall: Exit Poll Subgroup Instability

Exit polls are notoriously unreliable for subgroup analysis, especially for smaller demographic groups. A national exit poll with 15,000 respondents might contain only 1,200 Black respondents — a sample size where a 5-percentage-point shift from one election to the next is not statistically distinguishable from sampling noise. Claims about large shifts in Black, Hispanic, or Asian voter behavior that rest primarily on exit poll comparisons should be treated with great skepticism. Compare to pre-election polls, validated voter file data, and post-election surveys before concluding that a real shift has occurred.

Hispanic/Latino Voters: Heterogeneous and Complex

The analytical error most commonly made about Hispanic/Latino voters is treating them as a uniform bloc. Latino voters include:

• Cuban Americans in Florida, who have historically leaned Republican (though this pattern is evolving among younger Cuban Americans)
• Puerto Ricans in New York and the Northeast, who lean strongly Democratic
• Mexican Americans in California, Texas, and the Southwest, who lean Democratic but with real variation by generation and educational attainment
• Central American immigrants in the mid-Atlantic, whose political behavior is shaped by both immigration concerns and other economic priorities
• Dominican Americans in the urban Northeast, who lean Democratic
• Venezuelan Americans in South Florida, who have been moving Republican in response to anti-socialist messaging

These populations share a language heritage but are culturally, economically, and politically distinct. Aggregating them into a single "Latino vote" obscures more than it reveals. A strategy that works for Cuban Americans in Miami will not work for Mexican Americans in Albuquerque.

🌍 Global Perspective: Immigrant Political Incorporation

Political incorporation of immigrant populations follows patterns studied across many democracies. New immigrants are typically less politically engaged, partly due to registration and naturalization barriers and partly due to time and attention constraints. First-generation immigrants who do vote often prioritize immigration policy and economic opportunity. Second and subsequent generations are more likely to be shaped by the same dynamics as native-born citizens of similar socioeconomic backgrounds. Understanding which generation you're talking to is as important as understanding ethnic background.

📊 Real-World Application: Latino Voters in the Garza-Whitfield State

The Sun Belt state's 32% Hispanic/Latino population is not a uniform bloc, and Nadia knows this from her granular precinct data. The urban core's Hispanic population is primarily Mexican-American and Puerto Rican, with high Democratic inclinations. The agricultural border counties have a predominantly Mexican-American population with lower turnout and somewhat lower Democratic margins than their registration suggests. The growing exurbs have a newer Latino population with complex political tendencies.

Garza's ancestry — daughter of Mexican immigrants — gives her a potential authenticity advantage with Mexican-American voters, but her campaign avoids making this claim too explicitly, partly out of concern that it would seem like pandering and partly because their data shows that younger Latino voters in the state prioritize healthcare and economic issues over candidate identity. The "vote for someone who looks like you" appeal is more resonant among older Latino voters who remember a time when such representation was rare.

Asian American Voters: The Fastest Growing and Most Underanalyzed

Asian Americans are the fastest-growing racial group in the United States and increasingly numerous in Sun Belt states and suburban constituencies. They are also the most underanalyzed, partly because they are often treated as residual — "Other" — in demographic analyses and partly because the group is extraordinarily heterogeneous.

Asian Americans include communities whose political attitudes vary enormously: - Indian Americans, who have shifted strongly Democratic in recent cycles - Chinese Americans, who have been moving Democratic but with real variation - Vietnamese Americans, who have historically been more Republican (particularly older Vietnamese with connections to the anti-communist South Vietnamese government) - Korean Americans, who lean Democratic but with significant evangelical Christian communities that are more conservative - Filipino Americans, who lean Democratic - Japanese Americans, who have historically leaned Democratic

Aggregating these communities into "Asian American" produces a meaningless average. A research finding about "Asian American" voters tells you very little about Vietnamese-American voters in Orange County, Indian-American voters in Northern Virginia, or Chinese-American voters in the San Gabriel Valley.

⚖️ Ethical Analysis: The Invisibility of Small Groups

Survey research has a structural tendency to make smaller demographic groups invisible. When a poll has 600 respondents, the Asian American respondents might number 30-50 — a subsample too small for reliable analysis. Researchers respond by pooling groups ("minority voters") or simply ignoring them. But this invisibility has real political consequences: campaigns don't target communities they don't have data on, media don't cover electoral politics within communities they can't report on, and policymakers don't hear from groups that don't show up in their data. The decision about who is worth measuring is always, implicitly, a decision about who matters.

13.4 The Gender Gap: Origins, Drivers, and Magnitude

The gender gap in American politics — the consistent tendency for women to vote more Democratic than men — has been a durable feature of the electoral landscape since the 1980 presidential election, when it first became clearly visible. Understanding the gender gap requires distinguishing between what it is, what drives it, and how large it actually is.

Origins of the Gender Gap

Contrary to a common assumption, the gender gap did not emerge because women became more Democratic — it emerged primarily because men became more Republican. In the 1950s and 1960s, women were actually somewhat more Republican than men (partly due to higher religiosity and support for incumbents). Beginning in 1980, men's identification with and votes for Republicans increased, while women's partisan leanings remained relatively stable. The gap opened because the two groups moved apart.

The timing — 1980 — is significant. Ronald Reagan's campaign emphasized military strength, tax cuts, and opposition to the Equal Rights Amendment. These appeals were more effective with men than women, initiating the partisan divergence. Subsequent Republican positioning on social issues (particularly reproductive rights), welfare state programs, and conflict-oriented foreign policy has maintained and in some cycles deepened the gap.

What Drives the Gap Today?

Contemporary research identifies several distinct drivers:

Policy attitudes on social programs: Women are more likely than men to support government programs for healthcare, education, and social welfare — consistent with the historical pattern of gender differences in attitudes toward collective responsibility.

Reproductive rights: Since Roe v. Wade, abortion has been a gender-differentiated issue, with women more likely to prioritize reproductive autonomy as a voting consideration. This effect was amplified significantly after the Supreme Court's 2022 Dobbs decision.

Safety and violence concerns: Women are more likely to cite concerns about gun violence, domestic violence, and personal safety as political motivations — concerns where Democrats have historically taken stronger positions.

Economic security: Women are more economically vulnerable on average — higher rates of poverty, lower wages, greater dependence on social programs — and tend to favor policies that address economic insecurity more than men do.

Education levels: Because college-educated voters are moving Democratic and women are more likely to be college-educated (women now earn more bachelor's degrees than men), compositional change in educational attainment contributes to the gender gap.

The Gender Gap's Actual Magnitude

The gender gap is real but often overstated in popular discussion. In recent presidential elections, the gap in vote choice has been approximately 10-15 percentage points — significant, but not as dramatic as some narratives suggest. In 2020, women voted for Biden over Trump by approximately 57-42%; men voted for Trump over Biden by approximately 53-45% — a 15-point gender gap.

It is important not to lose sight of within-gender heterogeneity. The gender gap is largest among unmarried women, younger women, and college-educated women. Married women, older women, and non-college women are substantially less Democratic in their leanings — and married white non-college women actually lean Republican. "Women" is not a uniform political category.

Men's Politics: The Understudied Side

Most gender gap analysis focuses on women, but the movement of men — especially non-college men — toward Republicans is equally important and somewhat underanalyzed. Non-college men have moved dramatically toward Republicans over the past two decades, and this shift is not primarily explained by policy attitudes or economic interests. Research points to cultural identity factors: a sense that Democratic politics and institutions are aligned against masculine identity, traditional gender roles, and working-class male culture. This is a form of the social identity and symbolic politics dynamics we discussed in Chapter 11.

🔵 Debate: Is the Gender Gap Widening?

Since 2016, significant attention has been paid to the possibility that the gender gap is widening — particularly among younger Americans. Surveys suggest that younger women are substantially more liberal than their male counterparts, including on issues beyond the traditional gender gap drivers. Some researchers interpret this as a harbinger of long-term partisan divergence; others argue it reflects life-stage factors (young women in college and early careers; young men in different social positions) that will narrow as people age. The empirical jury is still out on whether this is a permanent widening or a temporary divergence.

13.5 Age and Generational Replacement

Age is one of the most robust predictors of political behavior, though its relationship with partisanship is more complicated than often portrayed. Two distinct dynamics operate simultaneously: life-cycle effects and generational effects.

Life-Cycle Effects

Life-cycle effects are changes in political behavior associated with aging rather than generational identity. The stereotyped version — "young people are liberal, old people are conservative, and everyone gets more conservative as they age" — is substantially wrong. Research shows that people do not reliably become more conservative as they age. What does happen:

• Turnout increases with age (up to a point) due to greater stakes, civic habit formation, and more stable residential patterns
• Party identification becomes more stable with age (the "impressionable years" effect — people form durable partisan attachments during young adulthood)
• Issue prioritization shifts — older voters care more about Social Security and Medicare; younger voters care more about student debt, housing, and climate

Generational Effects

Generational effects are differences in political behavior that reflect the political environments in which different cohorts came of age. The political experiences of formative years — roughly 18-25 — shape partisan identities that can persist for decades. The New Deal generation became Democratic because they came of age during FDR. The Silent Generation and early Boomers came of age in the postwar Republican era. Gen X came of age during Reagan. Millennials during 9/11, the Iraq War, and the 2008 financial crisis. Gen Z during Trump, COVID, and ongoing economic precarity.

Generational replacement — the slow process by which older cohorts die and younger cohorts enter the electorate — is one of the two mechanisms through which demographic change can produce political change. (The other is individual conversion — people changing their minds.) Generational replacement is slow but cumulative: if a new generation enters the electorate leaning Democratic, and an older Republican-leaning generation is dying, the net effect is gradual movement toward Democrats, even if no individual changes their mind.

The complication is that generational effects are not always stable. A generation's political identity formed in young adulthood can be modified by subsequent events. The Obama-coalition Millennials who voted Democratic in 2008 and 2012 showed less uniform Democratic loyalty in 2016 and 2020, particularly white Millennials without college degrees. And Gen Z voters have shown gender divergence that complicates generational analysis.

13.6 Religion and the Political Landscape

Religious identity has been one of the most powerful demographic predictors of American political behavior, and the religious landscape is changing rapidly.

The White Evangelical Alignment

The alliance between white evangelical Protestants and the Republican Party is one of the most durable coalitional relationships in contemporary American politics. Since the late 1970s — when Jerry Falwell Sr. and others organized the Moral Majority — white evangelicals have been overwhelmingly Republican in their presidential voting, typically giving Republican candidates 75-80% of their votes.

This alignment is primarily driven by cultural and social values: abortion, LGBTQ+ rights, religious freedom concerns, and cultural identity. Many white evangelicals feel that the Democratic Party is hostile to their values and lifestyle, making Republican identification a matter of group identity and self-defense as much as policy preference.

The Rise of the Nones

The fastest-growing religious category in the United States is "nones" — those who describe themselves as having no religious affiliation. This group, which has grown from about 5% of the population in the 1970s to roughly 28-30% today, leans strongly Democratic. Nones are disproportionately young, college-educated, and urban — overlapping with other Democratic-leaning demographic groups.

The growth of the nones is one of the most significant long-term demographic trends affecting partisan coalitions, because it suggests that the cultural and religious identity base of the Republican coalition may shrink over time. But this is exactly the kind of projection that can mislead: religiosity is also affected by life-stage factors (people often become more religious as they have children and age), and it is possible that today's young nones will look more conventionally religious in 30 years.

Catholic and Jewish Voters

Catholic voters were once a reliable Democratic constituency — the New Deal coalition included Irish, Italian, and Polish Catholic immigrants and their descendants. Over time, white Catholics have become much more politically mixed and now roughly mirror the national electorate. Latino Catholics remain more Democratic.

Jewish Americans remain a strongly Democratic group (typically 65-75% for Democratic presidential candidates), a pattern rooted in historical experiences of persecution and a strong cultural tradition of support for pluralism and civil rights. Recent Republican efforts to appeal to Jewish voters through support for Israel have generally not produced significant partisan movement, as most Jewish Americans prioritize domestic policy over Israel policy when choosing between parties.

13.7 Urban-Rural-Suburban Geography: The New Political Cleavage

As we touched on in Chapter 12, the urban-rural divide has become one of the most powerful demographic cleavages in American politics — arguably replacing class as the primary geographic axis of partisan competition.

The Dimensions of the Divide

The urban-rural cleavage is driven by multiple overlapping factors:

Density: Urban density is associated with Democratic voting even controlling for race, education, and income. There is something about dense, mixed-use urban environments that correlates with liberal political values — greater experience with diversity, more awareness of collective action problems, more reliance on public services.

Industrial structure: Urban economies are dominated by knowledge-sector work (tech, finance, healthcare, higher education); rural economies by agriculture, extraction, manufacturing, and logistical infrastructure. These different economic structures produce different political priorities.

Cultural identity: Urban and rural identities have become explicitly political in ways they weren't 30 years ago. "Urban" has become code for cosmopolitan, diverse, and liberal; "rural" for traditional, white, and conservative — even though actual urban and rural populations are more diverse than these stereotypes suggest.

Media ecosystem: Urban and rural communities consume different media, encounter different social networks, and have different lived experiences that produce divergent political realities.

The Suburbs: The Real Battleground

Between urban and rural lies the great American suburb — and within suburban areas lie the most genuinely competitive electoral territory in the country. As Chapter 12 discussed, suburban voters have been moving toward Democrats in recent cycles, particularly college-educated suburban women. But suburbs are extraordinarily heterogeneous:

Inner suburbs: Dense, diverse, often majority-minority, closely integrated with urban core economically. Trending strongly Democratic.

Middle suburbs: Traditional postwar suburban development, mixed demographic composition, significant variation by housing age and income. The genuine swing territory.

Outer suburbs / exurbs: Low-density, predominantly white, primarily car-dependent development from the 1990s-2010s. Trending Republican as non-college white voters and culturally conservative families have chosen these locations.

The suburban sorting process is ongoing and not complete. The same suburb can look meaningfully different in 2024 than it did in 2016 due to in-migration, demographic change, and political events.

13.8 Demographic Change vs. Political Change: The Critical Distinction

We have now surveyed the major demographic groups in the American electorate and their partisan tendencies. It is time to revisit the central caution of this chapter: demographic change does not automatically produce political change.

Several mechanisms can decouple demographic and political trends:

Turnout differentials: A demographic group can grow as a share of the eligible population but remain a stable or even shrinking share of actual voters if their turnout falls.

Within-group political change: Parties can make inroads with demographic groups that were previously more strongly aligned against them. If Republicans make gains with Hispanic voters, the growth of the Hispanic population may not translate into Democratic gains even if the demographic trend looks favorable.

Mobilization and counter-mobilization: Demographic change can motivate counter-mobilization by the groups that perceive themselves as losing demographic ground. Rising Latino populations in Sun Belt states have historically been associated with increased white conservative voter turnout — a "threat response" that partially offsets demographic gains for Democrats.

Party adaptation: Parties can adapt their platforms and coalition strategies to accommodate demographic change. The Republican Party's relative success in 2022 — better than many predictions — partly reflected effective outreach to specific Hispanic and Asian American communities.

🔴 Critical Thinking: The "Demographic Destiny" Fallacy

The emerging Democratic majority thesis failed not because its demographic projections were wrong but because it treated demographic groups as having fixed, permanent partisan preferences that could be projected forward. Political scientists call this kind of reasoning "variable determinism" — assuming that a correlation observed today is a fundamental law that will hold indefinitely. Demographic change creates electoral possibilities and sets constraints, but it does not determine outcomes. The political choices of parties, candidates, campaigns, and movements are the intervening variables that determine whether demographic possibilities are realized.

This fallacy runs in both directions. Just as some Democrats assumed demographic change would automatically produce their majority, some Republicans in the early Trump era assumed demographic change doomed them and that they needed to make dramatic policy changes. In fact, the Republican Party found ways to retain and expand its coalition without fundamentally changing its platform — primarily through non-college white voter mobilization and modest gains with specific minority communities.

13.9 ODA's Demographic Analysis: Adaeze Examines the Sun Belt Electorate

Adaeze Nwosu had been watching the Sun Belt states carefully for three years when the Garza-Whitfield race entered its final stretch. Her flagship analysis product — the ODA Electorate Composition Report — was designed precisely to address the demographic destiny fallacy by disaggregating population trends from voting trends.

"The question is never 'are there more of them,'" Adaeze told a foundation funder who wanted a briefing on the state's political future. "The question is always 'are more of them voting, and for whom?' Those are empirically different questions."

Sam Harding had built ODA's state demographic dashboard using the oda_voters.csv dataset — a synthetic but realistic representation of the state's voter file, combining registration records with census demographic data and past vote history. Their analysis disaggregated the electorate in ways that neither campaign's internal polling had fully done.

Their key findings:

The Latino turnout gap is the most important unresolved variable. The state's Latino citizen voting-age population was 34%, but their share of actual voters in recent cycles had been 26-28%. If that gap narrowed significantly — either through organic mobilization or organized effort — it would dramatically change the electoral math. Garza's campaign was investing in Latino outreach. But ODA's historical data suggested that previous "years of the Latino voter" predictions had been consistently optimistic.

Non-college Latino voters are cross-pressured. ODA's analysis showed that non-college Latino voters in the state held conservative views on immigration enforcement (many favor orderly but not draconian enforcement) and economic issues (small business ownership, skepticism of some union policies) alongside left-leaning views on healthcare and education. Neither party was speaking to the full complexity of this population.

The education gap within white voters is larger than the racial gap in some subregions. In the middle suburbs, the partisan gap between college-educated and non-college whites was larger than the gap between white and Latino voters. This meant that the "minority outreach" frame — while important — was too blunt for strategic resource allocation.

Young voter behavior is uncertain in both directions. Young voters (18-29) who came of age during the pandemic and its economic aftermath showed lower engagement with both parties and higher rates of third-party consideration in early surveys. This was simultaneously good news for Whitfield (some young voters who might vote Democratic were disengaged) and bad news (young voters who did turn out were slightly more likely to support Garza).

⚖️ Ethical Analysis: Who Gets Resourced

ODA's mission includes not just analyzing electorates but making analysis accessible to communities that are typically outside the campaign data ecosystem. Most campaign spending on demographic analysis focuses on identifying and mobilizing the communities that campaigns think will support them — which means that communities perceived as electorally marginal receive less outreach, less polling, and less data-driven attention. Adaeze has written publicly about how this creates a self-fulfilling prophecy: communities that are not targeted have lower turnout, confirming the campaigns' priors that they're not worth targeting. ODA's demographic tools are specifically designed to be usable by community organizations, not just campaigns, to break this cycle.

13.10 Nadia's Demographic Analysis of the Garza-Whitfield State

Inside the Garza campaign's analytics operation, Nadia Osei was running a parallel but differently motivated demographic analysis. Where ODA was trying to understand the electorate, Nadia was trying to win an election — and those goals require different analytical frames.

Her demographic model of the state electorate was built in tiers:

Tier 1 — Base voters: Black voters in urban core (high turnout, ~90% Garza), college-educated women in inner suburbs (high turnout, ~67% Garza), union households statewide (moderate turnout, ~58% Garza). These were not targets for persuasion; they were targets for mobilization — getting them to the polls.

Tier 2 — Soft support: Latino voters across the state (variable turnout, ~62% Garza on average but with huge internal variation), young voters in university towns (lower turnout, ~64% Garza), non-college Black men (moderate turnout, ~82% Garza — lower than Black women). These voters leaned toward Garza but needed some campaign attention to firm up.

Tier 3 — Persuadables: College-educated white men in middle suburbs (moderate-high turnout, ~48% Garza — barely competitive), non-college Latino men in exurbs (moderate turnout, ~51% Garza — flipping), moderate Republican women in outer suburbs (high turnout, ~40% Garza — needs to be higher). These were the campaign's persuasion targets.

Tier 4 — Unlikely converts: White non-college rural voters, strong Republicans statewide. Not worth campaign resources.

Nadia's model showed that Garza could win under several different scenarios: running up very large margins in the urban core plus holding her own in the suburbs, or achieving somewhat smaller urban margins but performing better with Tier 3 persuadables. The campaign's internal debate was about which scenario to pursue — with different implications for resource allocation, messaging, and candidate travel.

"The demographic analysis tells you where the votes are," Nadia said in a campaign strategy meeting. "It doesn't tell you which ones you can actually get. For that, you need the persuasion modeling." We'll develop that modeling in Chapter 29.

13.11 Measurement Shapes Reality: Who We Count and How

This chapter's material connects to Theme 1 — Measurement Shapes Reality — as well as Theme 2. The demographic categories we use to analyze electorates are not natural; they are constructed, contested, and consequential.

Race and ethnicity categories: The Census Bureau's race and ethnicity categories have changed repeatedly over the history of the census. The current system asks separately about ethnicity (Hispanic or not Hispanic) and race (with multiple options). Many respondents with mixed backgrounds choose "some other race" or leave questions blank. The political science research that treats racial categories as fixed and objective is working with data constructed by a bureaucratic and political process.

The "Hispanic" category: Hispanic or Latino is an ethnicity category, not a race category — which is why "Hispanic white" and "Hispanic Black" are both valid classifications. The political diversity within this category — as we've seen — is enormous. Treating Hispanic as a single political category is an analytical choice that serves certain purposes (tracking aggregate trends) but obscures others (within-group political diversity).

Gender binary limitations: Political survey research typically uses a binary gender measure (male/female) that excludes non-binary and gender non-conforming individuals, who represent roughly 1-2% of the population and may have distinct political behavior. The inability to measure and analyze this group is a form of erasure that serves data convenience over representational accuracy.

Rural-urban classification: The federal government and researchers use many different definitions of "urban" and "rural" — from the Census Bureau's urbanized areas to the USDA's rural-urban continuum codes. The political conclusions you reach about urban-rural political divergence depend substantially on which classification you use.

These measurement choices are not just academic. When campaign microtargeting is built on demographic data, the categories used determine who gets targeted and who gets ignored. When media coverage of "the Latino vote" treats it as uniform, it flattens the political diversity of 30+ distinct national-origin communities. When demographic destiny arguments are built on census projections, the classification choices embedded in those projections shape who counts as eligible to produce political change.

This is what "Who Gets Counted, Who Gets Heard" means at its deepest level: the construction of demographic categories in political data is an exercise of definitional power that has real consequences for political participation and representation.

Chapter Summary

The American electorate is changing in consequential ways: the education realignment has restructured white partisan coalitions along educational lines; the gender gap has grown, particularly among younger voters; racial and ethnic partisan coalitions show meaningful stability alongside important internal diversity and gradual change; generational replacement is slowly shifting the electorate's composition; religious change — particularly the growth of the nones — is restructuring the cultural coalitional landscape; and urban-rural-suburban geography has emerged as a dominant organizing cleavage.

These changes create real electoral possibilities — but not inevitable outcomes. The central lesson of this chapter is that demography is not destiny. Demographic trends are translated into electoral change through the filters of registration, mobilization, and turnout; through within-group political change and party adaptation; and through the choices of candidates and campaigns. The analyst who reads demographic trends directly into electoral projections is making a systematic error.

ODA's analysis of the Sun Belt electorate — and Nadia's campaign-level segmentation — illustrate the tension between understanding demography as civic fact and instrumentalizing it for electoral purposes. Both are legitimate, but they require different frameworks and impose different ethical obligations.

The chapters ahead build on this foundation: Chapter 14 examines turnout modeling in detail, applying the demographic insights here to the specific challenge of predicting who actually shows up. Chapter 16 returns to Python tools for visualizing the electoral landscape.

Key terms introduced: education realignment, gender gap, generational replacement, turnout-weighted demographics, demographic destiny, the Big Sort (in geographic context), exit poll subgroup instability, ecological fallacy, electoral composition, CVAP (citizen voting-age population)

Chapter 14 examines turnout modeling — translating the eligible electorate into the likely voter universe.

Extended Discussion: How Campaigns Actually Use Demographic Data

The theoretical frameworks developed in this chapter are foundational, but they can feel abstract until you see how practitioners translate them into day-to-day analytical work. This section traces the specific ways the Garza campaign's analytics operation uses demographic data — and where the theory meets the messy reality of a real campaign.

The Voter File as Demographic Infrastructure

Every serious modern campaign begins with a voter file — a database of registered voters compiled from state election authority records, augmented with commercial data and predictive modeling. In the Garza-Whitfield state, the Democratic Party maintains a statewide voter file through the party's data vendor, updated continuously with registration changes, vote history from each completed election, and a range of appended demographic and commercial data.

The basic voter file contains: name, address, date of birth, registration date, party registration (in states with party registration), and vote history in each election (whether voted — not for whom). Augmented versions add: estimated household income, home ownership, retail purchasing behavior, magazine subscriptions, organizational memberships, and, crucially, model scores.

The model scores — which are where much of the sophisticated demographic analysis lives — include: - Partisanship score: a probability from 0-100 estimating the likelihood the voter will vote Democratic if they turn out - Turnout score: a probability estimating the likelihood the voter will vote in this specific election - Persuadability score: a probability estimating how likely the voter is to change their vote intention in response to campaign contact - Issue-specific scores: probabilities estimating the voter's position on healthcare, immigration, climate, and other key issues

These scores are built from a combination of vote history, demographic characteristics, geographic context, and — where available — survey data linked to the voter file. The demographic analysis described throughout this chapter provides the theoretical backbone for how these scores are constructed.

The Demographic Paradox in Targeting

Here is a tension that Nadia Osei wrestles with constantly: the theoretical frameworks from this chapter tell her that demographic categories are heterogeneous, contested, and only probabilistically related to political behavior. But the targeting models she uses every day necessarily work through demographic categories — assigning probabilities based on race, education, age, gender, and geography.

The analyst's job is to hold both truths simultaneously: use demographic categories as practical tools for resource allocation, while maintaining intellectual humility about their limits and ethical awareness of their costs.

In practice, this means:

Using demographics as prior probabilities, not certainties. A voter coded as "Hispanic, non-college, male, age 35-44" gets a partisanship score that reflects the average behavior of people with those characteristics. But any individual in that demographic category may be quite different from the average. The targeting model works across thousands of contacts; it doesn't need to be right about any particular voter, just right enough in aggregate.

Updating with behavioral data when available. Prior surveys, door-knocking interactions, and phone banking conversations that yield a response update the individual voter's score more than their demographic profile does. A voter coded as "lean Republican" by demographics but who volunteers their support for Garza in a canvassing conversation gets reclassified immediately. The behavior trumps the demographic inference.

Disaggregating categories where data permits. In the Garza-Whitfield state, Nadia's team has disaggregated the "Latino" category into at least four sub-universes — urban Mexican-American, rural agricultural-belt Latino, suburban assimilated Latino, and new Latino immigrants — with different partisanship scores, turnout scores, and communication strategies for each. This isn't perfect disaggregation, but it's much better than treating 32% of the state as a single bloc.

Extended Analysis: The Education Realignment and Its Limits

The education realignment deserves more extended treatment than the main chapter sections provide, because it is both the most politically consequential current demographic trend and the one most subject to misinterpretation.

The Mechanism Question

Why has education become such a powerful predictor of partisan alignment among white voters? The simple answer — educated people are smarter and therefore less susceptible to Republican misinformation — is both condescending and empirically incorrect. Education does not produce uniform increases in political knowledge or sophistication; it produces specific types of knowledge and exposure to specific institutional cultures.

Several more defensible mechanisms have been proposed:

Cosmopolitanism and diversity exposure. College-educated individuals spend formative years in environments with greater demographic diversity, more explicit attention to inclusion and equity, and more international perspectives. These experiences are correlated with more liberal views on immigration, racial justice, and multiculturalism — which happen to align more with the Democratic platform.

Credential-economy vs. physical-economy attachment. The college premium — the wage advantage of degree-holders over non-degree holders — has grown substantially over the past four decades. College graduates have a material stake in an economy that rewards credentials and knowledge work, which aligns with Democrats' emphasis on education funding, technology investment, and institutional expertise. Non-college workers in physical and manual sectors have a different set of economic stakes.

Status anxiety and cultural threat. For non-college white workers, the same decades of credential-economy growth have been associated with wage stagnation, deindustrialization, and declining social status. The sense that the country's dominant institutions — universities, media, major corporations, federal agencies — no longer respect or represent their values creates openness to anti-establishment appeals. The Republican Party's pivot toward explicitly anti-elite, anti-expertise rhetoric under Trump matched this psychological demand.

Self-selection into educational institutions. People with liberal value predispositions may be more likely to pursue four-year degrees, producing a selection effect rather than (or in addition to) a socialization effect. Disentangling these mechanisms is genuinely difficult.

Geographic Variation in the Realignment

The education realignment is not geographically uniform. It is most pronounced in metropolitan areas — where college-educated and non-college residents live in the same communities, vote in the same elections, and compete in the same labor markets. In areas with very low educational attainment, there are too few college-educated residents to produce a visible education-based cleavage. In areas with very high educational attainment (university towns, tech hubs), the college/non-college distinction maps almost perfectly onto Democrat/Republican.

In the Garza-Whitfield state, the education cleavage is most visible in the middle suburbs surrounding the major metro area. The inner suburbs have high educational attainment and are strongly Democratic. The outer exurbs have lower educational attainment and are strongly Republican. The middle suburbs — the battleground — contain a mix, with the education-based cleavage clearly visible in precinct-level data. Precincts with more than 45% college graduates are lean-Democratic; precincts with less than 30% college graduates are lean-Republican; precincts in between are competitive.

Non-College Voters Are Not All the Same

One of the analytic errors that the education realignment narrative has produced is treating "non-college voters" as a monolithic Republican-leaning group. Non-college voters are actually the most demographically diverse segment of the electorate — they include Black non-college workers (who remain strongly Democratic), Hispanic non-college workers (who are more mixed and moving), Asian non-college workers (extremely diverse), and white non-college workers (who have moved dramatically Republican).

When analysts or media commentators talk about "non-college voters moving Republican," they almost always mean non-college white voters. The aggregate non-college category includes groups with very different partisan trajectories. Collapsing them obscures more than it reveals.

This is the ecological fallacy problem applied to the education realignment: using an aggregate pattern (non-college voters overall have shifted somewhat toward Republicans) to make inferences about the behavior of specific subgroups within that category. The direction and magnitude of the shift varies enormously by race, and treating the aggregate as representative of any subgroup is an error.

Religion, Secularization, and the New Cultural Cleavage

The religious transformation of the American electorate interacts with education and generation in ways that make it one of the most complex demographic stories currently unfolding. Let us trace the pattern more carefully than the main chapter sections permitted.

The White Evangelical Alignment — How Durable Is It?

White evangelical Protestantism's alliance with the Republican Party is not historically inevitable — it was constructed, primarily in the 1970s and 1980s, through deliberate political organizing. Before the late 1970s, Southern Baptist Convention (the largest evangelical denomination) leaders often maintained Democratic Party ties as part of the broader Southern Democratic tradition. The mobilization of evangelical Christians into the Republican coalition by figures like Jerry Falwell Sr. and the Moral Majority was a strategic achievement, not a natural alignment.

The question is whether this alignment will persist as the evangelical population itself changes. Younger evangelicals (under 40) show somewhat more heterodox views on climate, racial justice, and economic inequality than older evangelicals — areas where their views are less compatible with the current Republican platform. Black evangelicals, who share theological conservatism with white evangelicals, remain strongly Democratic. The "evangelical" label itself is becoming contested, with some younger Christian conservatives preferring "post-evangelical" or simply "Christian" over a label they associate with partisan politics.

None of this means the evangelical-Republican alignment is about to collapse. Generational change in religious identity is slow, and the specific issues that cemented the alliance — reproductive rights, LGBTQ+ rights, religious freedom in schools and public life — remain salient and politically divergent. But the alliance may be somewhat less stable in 2030 than in 2010, particularly among younger evangelicals who did not come of age in the Moral Majority era.

The "Nones" Are Not a Simple Democratic Coalition

The rapid growth of religious non-affiliation — from roughly 5% of the U.S. adult population in the 1970s to 28-30% today — is one of the most dramatic cultural changes in recent American history. And the nones do lean Democratic, particularly on social and cultural issues.

But the nones are not monolithically progressive. They include:

• Secular humanists and atheists, who tend to be strongly Democratic and engaged on church-state issues
• "Spiritual but not religious" individuals, who may hold idiosyncratic political views not well captured by either party platform
• Nominally Christian lapsed churchgoers, who grew up in religious households but have drifted away without adopting a secular identity — this is the largest subgroup and is politically moderate to conservative on many issues
• Young adults who never affiliated, a growing cohort in the youngest age groups who are politically liberal on social issues but often skeptical of party institutions generally

The political behavior of the "nones" is therefore heterogeneous in ways that the aggregate Democratic lean obscures. For campaign targeting purposes, a self-described "none" who is a secular humanist in an urban area is a very different voter than a "none" who is a lapsed Catholic in a rural county.

Data Justice and the Politics of Who Gets Measured

We have argued throughout this chapter that demographic data is never politically neutral — that the construction of categories has consequences for who is represented and who is not. Let us make this concrete with a specific example that Adaeze Nwosu has used in ODA's public presentations.

The Case of Native American Voters

Native American voters — roughly 1% of the U.S. population but a higher percentage in states like Montana, North Dakota, South Dakota, and Oklahoma — are among the most systematically under-measured groups in political survey research. Several factors contribute:

Geographic concentration in hard-to-survey areas: Many Native American communities are on reservations, which are geographically remote, have lower landline telephone penetration, and present logistical challenges for face-to-face survey research. Online panels undersample reservation populations due to lower internet access rates.

Small population size: A national survey with 1,200 respondents will typically contain only 12-15 Native American respondents — far too few for reliable subgroup analysis. State-level surveys in states with large Native American populations do better, but resources for such surveys are limited.

Cultural and historical barriers: Historical experiences with federal data collection — from Census Bureau counts to Indian Health Service data — have created well-founded wariness about data sharing in some communities. Survey response rates among Native Americans are often lower than among other groups.

Category complexity: The "American Indian/Alaska Native" Census category encompasses more than 570 federally recognized tribes with distinct languages, cultures, governance structures, and histories. Political behavior varies substantially across communities in ways that the single category cannot capture.

The practical consequence: campaigns rarely have reliable data on Native American voters, rarely target them with specific outreach, and consequently often undercount their potential contribution to electoral outcomes. In North Dakota's 2018 Senate race, Native American turnout — energized by concerns about voting ID laws and treaty rights — almost certainly contributed to incumbent Heidi Heitkamp's close (if ultimately losing) performance. The margin between what polls predicted and what happened was partly explained by Native American voters who weren't appearing in survey samples.

This example illustrates the "Who Gets Counted, Who Gets Heard" theme at its most concrete: systematic measurement failures create systematic representation failures, which create systematic policy failures in democratic systems that are supposed to respond to the concerns of all citizens.

ODA's Response

Adaeze's framework for addressing data justice in demographic analysis has three components, which Sam Harding has implemented in ODA's tools:

Visible uncertainty: Rather than suppressing estimates for small groups or presenting them as if they were as reliable as large-group estimates, ODA's tools display explicit confidence intervals for all demographic subgroup estimates. A category with 45 survey respondents shows a very wide confidence interval. This visual representation of uncertainty makes the data quality differences visible rather than hidden.

Community-partnered data collection: For the groups most systematically underrepresented in standard surveys — Native Americans, rural Asian Americans, recent immigrants, undocumented residents — ODA has developed partnerships with community-based organizations to conduct targeted supplemental surveys. These surveys are smaller but more representative of their specific communities than the fragments available from national samples.

The "missing data is data" principle: When a group is consistently absent from survey datasets, ODA treats that absence as information — evidence of structural barriers to participation — rather than as a reason to exclude the group from analysis. Reports include sections on groups for whom data is unavailable or unreliable, explaining what is known about structural barriers and what the implications of the data gap might be for electoral and policy outcomes.

These practices don't fully solve the data justice problem — that would require substantially more resources, political will, and structural change in how political data infrastructure is funded and organized. But they represent one nonprofit's attempt to address the gap between the data that campaigns and analysts work with and the democratic ideal of a fully represented polity.

The Demographic Composition of the Garza-Whitfield State: A Complete Picture

We have examined individual demographic dimensions throughout this chapter. Now let us put them together into an integrated picture of the Garza-Whitfield state's electoral landscape.

The state's CVAP is approximately: - 38% white non-Hispanic - 32% Hispanic/Latino - 18% Black - 12% Asian American and other

But the likely voter electorate — filtered for registration, predicted turnout, and election-specific mobilization — looks considerably different: - Approximately 44% white non-Hispanic (overrepresented relative to CVAP) - Approximately 26-28% Hispanic/Latino (significantly underrepresented) - Approximately 20-21% Black (slightly overrepresented relative to CVAP due to high turnout rates) - Approximately 8-10% Asian American and other (slightly underrepresented)

This transformation from CVAP to likely voter electorate is the first and most important calculation in any electoral analysis of this state. The state is majority-minority in its citizen adult population but has a plurality white electorate among likely voters. This gap — produced by the differential turnout and registration effects described throughout this chapter — is the most consequential single fact about the state's electoral structure.

When demographic destiny arguments suggest that the state's majority-minority population will inevitably produce a Democratic majority, they are reasoning from CVAP. When Whitfield's campaign argues that the "real voters" in this state lean his direction, they are implicitly reasoning from the likely voter electorate. Both are using real numbers selectively. The politically important truth is that the gap between the two is large, partially malleable through mobilization, and the primary site of electoral contestation.

Nadia's job — and the job of every campaign in a diverse, competitive state — is to close that gap on her side while preventing the other side from doing the same. That is the practical meaning of "Who Gets Counted, Who Gets Heard" at the campaign level. And it is why demographic analysis is never just an academic exercise: it is a political act, with consequences for who exercises power and who does not.