Chapter 12: Partisanship, Polarization, and Sorting

Opening: The Map That Lied

Pull up the famous county-level election map from any presidential election in the last three cycles. A vast sea of red with islands of blue — the coastal cities, the major metro areas, and a scattered few university towns. Many people look at that map and conclude that America is overwhelmingly Republican. Conservatives sometimes brandish it as proof. Liberals sometimes despair at it.

The map is not lying exactly, but it is profoundly misleading. The red counties contain fewer people. The blue counties contain more. And the intermediate shades — the purple zones where real competition happens — look smaller than they are because geography doesn't weight votes. Wyoming is three times larger than Connecticut on a map. Connecticut has about twice as many people.

This is Theme 5 in its most literal form: the gap between the map and the territory. The visual representation of geographic sorting tells you something important about how American partisanship has become spatially concentrated, but it badly distorts the actual distribution of political preferences in the electorate. Understanding both what the map shows (real geographic clustering) and what it obscures (a much more competitive underlying electorate) is essential for any serious political analyst.

This chapter examines one of the most consequential transformations in American politics over the past fifty years: the restructuring of partisanship through sorting, polarization, and nationalization. These are related but distinct phenomena, and conflating them leads to systematic analytical errors. We'll work through each carefully, then connect them to their implications for polling, analysis, and the Garza-Whitfield race.

12.1 Understanding What Changed: A Historical Baseline

To appreciate how much the partisan landscape has changed, you need a baseline. Start with the 1950s and 1960s American party system that the Michigan scholars studied.

The Democratic Party of 1960 contained within it a genuinely extraordinary coalition: Northern liberals, big-city machine politicians, organized labor leaders, Southern segregationists, and Black voters in Northern cities who voted Democratic despite the party's Southern wing's brutal enforcement of Jim Crow. The Republican Party was similarly heterogeneous: moderate Northeastern WASP establishment figures, Midwestern conservative businessmen, and a scattering of Black voters whose families had supported Lincoln's party for generations.

These coalitions look incoherent from our current vantage point because they were ideologically incoherent. Conservative Democrats and liberal Republicans existed in large numbers. Southern Democrats voted with Republicans on social issues. Northern Republicans voted with Democrats on some economic issues. Floor coalitions in Congress shifted constantly based on the issue at hand.

The parties were what political scientists call "non-sorted" — there was no strong alignment between ideology and party membership. A voter could be conservative and Democratic, liberal and Republican, without obvious contradiction. Party identification was a regional and social identity as much as an ideological one.

The Transformation

Between the 1960s and the 2020s, the parties underwent a fundamental realignment. The trigger was the Civil Rights Act of 1964 and the Voting Rights Act of 1965, which President Johnson signed knowing — in his famous phrase — that he was giving the South to the Republicans for a generation. He was right, though it took longer than a generation. The conservative white South moved steadily Republican through the 1970s, 1980s, and 1990s; liberal Northeastern Republicans moved steadily Democratic.

By the 2020s, the partisan landscape looked completely different:

• The percentage of liberal Republicans (measured by ideology questions) had fallen from roughly 30% in the 1970s to under 5%
• The percentage of conservative Democrats had fallen similarly dramatically
• The correlation between party identification and ideology (liberal-conservative self-placement) had risen from about r = 0.35 in 1960 to r = 0.65+ today
• In Congress, ideological overlap between the parties — once substantial — had virtually disappeared

📊 Real-World Application: Seeing Sorting in Data

The ANES time series, which has asked respondents about both party identification and ideology since 1972, is the best single data source for tracking partisan sorting over time. If you run a simple crosstab of party ID by ideological self-identification from the 1972 survey versus the 2020 survey, the difference is stark. In 1972, 25% of self-identified Democrats called themselves conservative; in 2020, that figure was about 11%. In 1972, 22% of self-identified Republicans called themselves liberal; in 2020, less than 5%. The parties have become ideologically homogeneous in ways they never were before.

12.2 Sorting vs. Polarization: A Critical Distinction

Before going further, we need to establish a distinction that trips up many analysts and almost all political journalists: sorting and polarization are not the same thing, and confusing them leads to fundamentally different — and often wrong — conclusions.

What Sorting Means

Partisan sorting is the process by which ideological liberals have become Democrats and ideological conservatives have become Republicans. The parties are now sorted along the ideological spectrum: if you know where someone falls on a liberal-conservative dimension, you can predict their party ID much better than you could 50 years ago.

Sorting does not necessarily mean that the distribution of opinion in the public has shifted. If there were 40% liberals, 40% conservatives, and 20% moderates in 1970 and there are still 40% liberals, 40% conservatives, and 20% moderates today, but all the liberals are now Democrats and all the conservatives are now Republicans, that is sorting without polarization. The parties have become more ideologically distinct, but the electorate as a whole is no more extreme.

What Polarization Means

Ideological polarization is the actual movement of public opinion toward the extremes. It implies that people who were once moderate have become more conservative or more liberal — that the distribution of opinion in the electorate has shifted from a hump in the middle (a normal distribution centered at moderate) to a bimodal distribution with peaks at liberal and conservative.

The empirical evidence here is more mixed than popular commentary suggests. Some issues show genuine mass polarization — abortion attitudes, for instance, have moved in opposite directions among Democrats and Republicans since Roe v. Wade. Gun control attitudes have polarized. But on many other issues, public opinion has remained relatively stable even as partisan conflict has intensified. Most Americans still favor some form of gun background checks, support some role for government in healthcare, and hold immigration opinions that don't fit neatly at the extremes.

This finding — that the parties have polarized more than the public — reflects the particular intensity of elite polarization: elected officials, party activists, major donors, and media figures have moved to the extremes faster and further than ordinary voters. When voters sort into the two parties, they become associated with elites who are more extreme than they are.

Affective Polarization: A Different Dimension Entirely

The third concept is the most politically important of the three: affective polarization. This is not about ideology at all — it is about feelings. Affective polarization refers to the growing emotional gap between how people feel about members of their own party (warm, favorable) versus members of the opposing party (cold, hostile, contemptuous).

The evidence for affective polarization is compelling and alarming. On the ANES feeling thermometer (where 100 is warm/favorable and 0 is cold/hostile), average ratings of the in-party have remained roughly stable, while average ratings of the out-party have fallen dramatically — from around 40-45 in the 1980s to under 25 in recent years. Americans have not changed how much they like their fellow partisans; they have become dramatically more hostile to the other side.

Affective polarization has grown even among voters who are not themselves ideologically extreme. A moderate Democrat may hold genuinely middle-of-the-road policy views but still feel intense hostility toward Republicans as a group — and vice versa. This is the implication of social identity theory applied to partisan identity: the mere fact of categorization into parties is sufficient to generate in-group preference and out-group hostility, independent of actual ideological differences.

🔴 Critical Thinking: Is Affective Polarization a Problem for Democracy?

Affective polarization raises difficult questions about democratic governance. Democratic theory requires some minimal willingness to accept electoral defeat and recognize the legitimacy of opponents. When partisan hostility is high, losers are more likely to question the legitimacy of elections, and winners are more likely to use power to harm the other party's supporters. Some scholars argue that affective polarization is the primary threat to democratic norms in contemporary American politics — not because voters have become more extreme ideologically, but because they have become more tribally hostile. Others argue that intense partisan conflict is a normal feature of competitive democracy and that the real problem is not polarization but the failure of institutions to channel it productively. Where do you land?

12.3 Negative Partisanship and the Hollow Middle

One of the most important consequences of high affective polarization is the emergence of what Alan Abramowitz and Steven Webster call negative partisanship: the primary motivation for many voters is not support for their own party but intense opposition to the other party.

In a world of negative partisanship, the traditional assumption that "independents" are persuadable moderates breaks down. Many self-described independents — particularly "independent leaners" — are actually strong negative partisans who reliably vote for one party not because they love it but because they despise the other. Behavioral analysis of ANES data shows that independent leaners vote for their favored party at nearly the same rate as weak partisans.

This has profound implications for campaign strategy. If you're Jake Rourke managing Whitfield's campaign, you might initially target "soft Republicans" and "independent leaners" as your persuasion universe. But if many of those leaners are negative partisans who vote Republican regardless of what Whitfield does — because the alternative is unthinkable — then spending resources persuading them is wasteful. The truly persuadable universe is smaller and weirder than it looks: it consists of voters who are genuinely indifferent between the parties, often because they're cross-pressured on different issues or genuinely skeptical of both parties.

💡 Intuition: The Shrinking Middle

As affective polarization increases and negative partisanship becomes more common, the genuinely persuadable electorate shrinks — but the number of people who describe themselves as "independent" may not shrink at all, because social desirability norms make "independent" a respectable self-description even for highly partisan voters. This creates a persistent measurement problem: independence in identification does not mean independence in behavior. The analyst who takes "independent" self-identification at face value will overestimate the persuadable electorate.

12.4 The Nationalization of Elections

Closely related to sorting is one of the most consequential trends in American electoral politics: the nationalization of elections. Before the 1990s, congressional and Senate elections had a significant idiosyncratic, local component. The slogan "All politics is local" — associated with Tip O'Neill — captured a real phenomenon: incumbency mattered enormously, local name recognition counted, and split-ticket voting was common.

Today, all of that has changed. The correlation between presidential vote share and Senate or House vote share in the same constituency has risen dramatically. Voters who would have once split their tickets — voting for a popular local senator of one party while supporting the other party's presidential candidate — now rarely do so. The national partisan tide has become nearly sovereign.

Why Nationalization Happened

Several forces drove nationalization:

Sorting of elected officials: As conservative Southern Democrats and liberal Northeastern Republicans disappeared from Congress, voters lost the cognitive category of "a Democrat I might vote for even though I'm Republican." The parties became more uniformly associated with national positions.

Media change: The collapse of local news and the rise of national media — cable news, social media, online political information — means voters receive more national political cues and fewer local ones. A decade ago, a local senator might have been known primarily through local news coverage of their constituent services. Today, they are evaluated primarily through national partisan media frames.

Partisan cue-taking: As the parties have sorted, the party label itself carries more information about what a candidate believes. Voters who don't know much about a specific senator can reasonably infer their positions from the party label alone. This reduces the premium on local name recognition and personal vote.

Campaign finance nationalization: National party organizations and outside groups now play larger roles in Senate and House races, bringing national messages and national money that crowd out local and candidate-specific communication.

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

Nadia Osei's data shows a striking pattern: in the previous Senate election in this state, roughly 12% of voters split their ticket between the presidential and Senate candidates. Based on current survey data, she expects only about 6-7% of voters to split this cycle. The nationalized environment is both a constraint and an opportunity. The constraint: Garza will rise and fall partly with the national Democratic brand, regardless of her individual qualities. The opportunity: national energy and enthusiasm — small-dollar fundraising, volunteer mobilization, media attention — is more available to competitive Senate candidates than it was 20 years ago precisely because nationalization has made Senate races feel important to national activists.

Down-Ballot Coattails in a Nationalized Era

The coattail effect — presidential candidates pulling down-ballot members of their party to victory — has always existed. But nationalization has generalized and routinized it. In earlier eras, coattails were episodic: a strong presidential candidate might help a few weak senators in marginal states. Today, the national partisan tide operates more continuously. House and Senate candidates in competitive districts now track presidential approval numbers closely not just at election time but throughout the two-year cycle.

For state legislative races, the nationalization effect is especially striking. State legislators who once won by 15 points based on local reputation and constituent service now face stiff competition from candidates who barely campaign, based purely on national partisan alignment. This has had consequences for candidate quality — the incentive to invest in a state legislative career is reduced when survival depends primarily on national forces beyond any candidate's control.

🔗 Connection: Nationalization and Chapter 22

We'll examine down-ballot dynamics in much greater depth in Chapter 22, particularly how analysts approach races where the local component of vote choice is shrinking. For now, note that nationalization doesn't mean local factors are irrelevant — they can still be decisive at the margins in competitive districts — but they no longer operate independently of national forces.

12.5 The Big Sort: Geographic Sorting and Its Consequences

In 2008, journalist Bill Bishop published The Big Sort, arguing that Americans were increasingly clustering into ideologically homogeneous geographic communities — choosing to live near people who shared their political views. The mechanism was not primarily political: people don't usually choose a neighborhood because it votes the right way. But the lifestyle and value preferences that correlate with politics — urban density vs. suburban space, cultural amenities vs. outdoor recreation, demographic diversity vs. homogeneity — are reflected in residential choices that create politically sorted geographies.

The empirical record on the Big Sort is more complicated than Bishop's argument suggested. Geographic polarization is real and growing, but much of it reflects compositional effects: Democrats have moved to cities faster than Republicans, and cities have grown relative to rural areas, rather than existing communities becoming more uniform. The sorting is partly a sorting of people into different types of places rather than existing places becoming more politically uniform.

Still, the consequences of geographic sorting are real and important:

Winner-take-all in most geographies: In a sorted geography, most individual precincts and counties are not competitive at all. Competition happens at the margin — in a small number of districts and states that are near the partisan median.

Reduced cross-partisan contact: If you live in a politically sorted community, you're less likely to have close relationships with people who hold opposing views. Research suggests this reduces empathy toward political opponents and makes affective polarization worse.

Distorted geographic representation: The geographic concentration of Democrats in urban areas means that Democratic votes are "wasted" in winner-take-all elections at a higher rate than Republican votes, which are more efficiently distributed across rural areas and smaller cities. This produces a persistent Republican lean in the Electoral College and the House that doesn't reflect the national vote distribution.

Media market geography: The Big Sort means that local media markets are increasingly partisan. A TV station in rural Georgia covers political news for an almost entirely Republican audience; a station in Atlanta's urban core covers politics for an almost entirely Democratic one. This shapes what gets covered and how.

⚠️ Common Pitfall: Confusing Geographic Sorting with Voter Extremism

The Big Sort is often interpreted as evidence that voters have become extreme — that sorted communities are full of ideological zealots. This is largely wrong. Sorted communities are often composed of normal people with fairly conventional views who happen to share the same basic partisan orientation. The political extremism associated with these areas comes more from the lack of competitive feedback — when one party always wins, primary elections become the only real contest, which selects for ideologically purer candidates — than from voter extremism per se.

12.6 How Polarization Affects Polling

For political analysts, polarization creates several systematic challenges in survey research that go beyond standard sampling error.

Partisan Nonresponse Bias

In Chapter 9, we discussed the general problem of nonresponse. Polarization creates a specific version of this: partisan differential nonresponse, sometimes called "mode-specific partisan nonresponse" or more loosely "poll shyness." The phenomenon works like this: when one party's voters are more enthusiastic about participating in polls — because they feel the political environment favors them, because surveys feel like a validation of their views, or because they're more civically engaged in general — surveys will oversample that party's voters even with identical sampling procedures.

The practical consequence is that poll averages can show systematic bias correlated with the political environment. When Republicans are enthusiastic and Democrats are demoralized, Republicans oversample in polls, and polls show an apparent Republican lead that overstates their actual position. When Democrats are energized — as in wave years — the opposite occurs.

The 2020 presidential election saw striking differential nonresponse: polls systematically overstated Biden's lead in key battleground states, particularly among non-college white voters (who leaned Republican and were underrepresented in poll samples). The 2022 midterms showed similar patterns. Identifying and correcting for differential nonresponse is one of the hardest problems in contemporary polling.

📊 Real-World Application: Meridian's Nonresponse Modeling

Dr. Vivian Park at Meridian Research has spent considerable effort developing a partisan nonresponse correction methodology. Her approach: use prior election results at the geographic level to calibrate the expected partisan distribution of respondents, then weight current surveys to match that expected distribution rather than simply demographic targets. The challenge is that the expected partisan distribution is itself uncertain — you don't know if there's been real partisan change until after the election. This circularity is what makes differential nonresponse so treacherous: the correction requires assumptions about what you're trying to measure.

Social Desirability Bias in a Polarized Environment

Social desirability bias — the tendency to give socially acceptable answers rather than honest ones — is a persistent challenge in survey research. In a polarized environment, it takes on particular forms.

When one candidate is seen as more socially acceptable by survey respondents, honest supporters of the other candidate may underreport their preference. This is the "shy Tory" or "hidden Trump voter" hypothesis: Republican-leaning voters in some contexts may underreport their support for Republican candidates because they perceive the social cost of admitting that preference to a stranger on the phone to be non-trivial.

The evidence for this effect is mixed. Some analysts point to persistent understating of Republican performance relative to polls as evidence of a "shy Republican" effect. Skeptics argue that partisan nonresponse is a better explanation for the same patterns. Disentangling these explanations is genuinely difficult because they make similar predictions about poll bias but require different corrections.

Social desirability can also run in the other direction: in strongly partisan communities, expressing support for the other party carries social costs, and voters may overstate support for the locally dominant party. In a Republican-dominated rural county, a genuine Biden voter might describe themselves as undecided rather than face social friction. In a heavily Democratic urban precinct, a genuine Trump voter might do the same.

The "Herding" Problem Under Polarization

Polarization creates another polling pathology: herding, where pollsters consciously or unconsciously adjust their results to be closer to the existing consensus of public polls. This happens because in a polarized environment, a poll that shows results far from the consensus is immediately contested and criticized — it becomes a story about the pollster's methodology rather than the race. The social and professional costs of being an outlier are high.

The result is that the distribution of public polls around the true value may be artificially compressed, because outliers that would reveal the spread of true uncertainty are suppressed. This makes poll aggregation seem more reliable than it is — all the polls look similar, so the average seems robust, but the underlying estimates may all share a common bias that the clustering has hidden.

🔵 Debate: How Much Should Pollsters Adjust for Known Biases?

Pollsters face a genuine dilemma: if they identify a systematic bias (like differential partisan nonresponse) and correct for it, they are making a modeling assumption that might be right or wrong. If they don't correct, they publish results they know may be biased. The professional norm has been to correct for observable demographics but not to make explicit partisan corrections, because the latter involves assumptions about "what the electorate should look like" that can shade into political judgment. More recently, some pollsters have begun experimenting with partisan weights — explicitly weighting their samples to match the expected partisan distribution. This is controversial. Where should the line between technical correction and motivated reasoning be drawn?

12.7 Populism and Partisan Realignment

The era of partisan sorting and polarization has also been an era of partisan realignment — a genuine restructuring of which social groups belong to which party. The most significant ongoing realignment is the class realignment: as we'll see in greater depth in Chapter 13, college-educated voters have been moving toward Democrats while non-college voters have moved toward Republicans, cutting across prior coalitional patterns.

This realignment has facilitated the rise of populist rhetoric — particularly right-wing economic populism — as a force within the Republican Party. Tom Whitfield's candidacy in our Garza-Whitfield race is a product of this realignment: he is a business owner running on economic populism against elites of both parties, appealing to non-college white voters and some non-college minority voters who have responded to economic anxiety with anti-establishment sentiment.

The connection between sorting, polarization, and populism runs through a common mechanism: when parties are sorted, elites and non-elites within each party share a clear partisan identity but may have divergent economic interests. Non-college Republican voters may benefit from higher minimum wages or trade protections that their college-educated Republican co-partisans oppose. This creates within-party tension that populist candidates can exploit by claiming to represent "the people" of the party against its elite faction.

🔗 Connection: Preview of Chapter 34

Chapter 34 will offer a full treatment of populism as a political phenomenon and an analytical challenge. For now, note the structural connection: partisan sorting helps explain why populist energy within the Republican Party is directed at Republican elites rather than at Democrats — these are people who identify as Republican but feel alienated from what Republican governance has delivered.

12.8 Geographic Sorting and the Garza-Whitfield State

The purple Sun Belt state in which Garza and Whitfield are competing is an interesting case study in geographic sorting precisely because it hasn't fully sorted yet. The state contains significant urban growth — a major metro area that has become substantially more Democratic over the past decade — and a large rural interior that has become substantially more Republican. The suburbs in between are the true battleground.

Nadia Osei's geographic analysis of the state reveals a pattern familiar to analysts in Arizona, Georgia, and Nevada: the overall D+2 registration advantage masks dramatic within-state heterogeneity. The three largest urban counties have Democratic advantages of 15-25 points. Twelve rural counties have Republican advantages exceeding 30 points. The twelve suburban counties surrounding the major metro area range from D+3 to R+8. It is these suburban counties that will determine the election.

The Big Sort dynamics play out in specific ways in this context:

Urban growth advantage for Democrats: In-migration to the major metro area has skewed Democratic, partly because the people attracted to a growing tech-and-health services economy tend to be college-educated and younger — both characteristics that have been moving Democratic in the realignment.

Rural entrenchment for Republicans: The rural interior has become more uniformly Republican over the past decade, reducing the ticket-splitting that once allowed Democrats to compete there. Even Democratic candidates who were once competitive in these areas now face structural headwinds.

Suburban volatility: The inner suburbs are in genuine transition. Long-time residents who moved to the suburbs in the 1990s and 2000s, when suburb = Republican was a safe assumption, are now politically mixed. New residents are younger and more Democratic. The result is a rapidly changing partisan composition that makes polling difficult: the electorate in a competitive suburban county looks meaningfully different in 2024 than it did in 2020 or 2018.

Jake Rourke on the Whitfield side is banking on two things: turning out the enthusiastic rural base at very high rates, and making inroads with the non-college suburban voters who lean Republican on economic issues but may be put off by some of the more strident Republican positions. Nadia's strategy is the mirror image: run up the margin in the urban core and hold on in the suburbs.

12.9 Sorting and Its Effects on Analytical Practice

The era of sorting and polarization has created specific challenges and opportunities for the political analyst that didn't exist in earlier, less sorted political environments.

Cross-Validation Across Demographics is Harder

In earlier, less sorted eras, demographic subgroups could be used as rough cross-checks: if your poll showed 95% of Black voters supporting the Democrat when recent elections had shown 88%, that discrepancy was a warning sign. In a highly sorted environment where within-group variation is lower, subgroup cross-validation becomes both easier (the expected distributions are more predictable) and more dangerous (small departures from expectations carry more information but are also harder to distinguish from sampling noise in small subgroups).

The Problem of the "Missing" Moderate

Sorted parties with high affective polarization have made it increasingly difficult to identify genuine moderates. Self-described "moderates" in survey research are a heterogeneous group: some are genuine centrists with mixed views across issues; some are "moderate" in affect (not very interested in politics, not very partisan) while holding extreme positions on specific issues; some use "moderate" as a respectable social self-description while voting consistently for one party. The analyst who treats all self-described moderates as persuadables is making a systematic error.

Tracking Electoral Nationalization Empirically

One practically useful exercise is to measure the degree of nationalization in a specific state or district by correlating recent presidential vote share with recent Senate or House vote share at the precinct or county level. High correlation (r > 0.90) indicates that national forces are dominant; lower correlation indicates that local factors still play a meaningful role. In the Garza-Whitfield state, the precinct-level correlation between presidential and Senate vote share has risen from about r = 0.78 in 2012 to r = 0.88 in 2020 — a significant increase in nationalization, but still enough local variation to give both campaigns reason to believe candidate-specific factors can matter.

🧪 Try This: Measuring Partisan Sorting in Your Own State

Using publicly available ANES state-level data or your state's exit poll data from multiple election cycles, calculate the correlation between party identification and ideological self-placement at three time points: 1990, 2008, 2020. How much has the correlation changed? What does the direction and magnitude of change tell you about sorting in your state's electorate?

12.10 Both Campaigns Navigate a Polarized State

The polarization and sorting dynamics described in this chapter create specific strategic constraints for both campaigns in the Garza-Whitfield race.

For Garza's campaign: High affective polarization means her base is enthusiastic and energized — negative partisanship is working for her in the sense that Democrats are highly motivated to prevent a Whitfield victory. The risk is that affective polarization also makes it harder to reach the voters she actually needs: suburban independents and soft Republicans who are put off by partisan intensity. Nadia's data shows that the most persuadable voters in the state are the ones most likely to be turned off by strident partisan messaging. The campaign must somehow mobilize the base (who need intensity) and reassure persuadables (who are spooked by intensity) at the same time.

For Whitfield's campaign: His populist appeal leverages the within-party tension created by sorting — he is running against Republican elites as much as Democrats, which is a coherent strategy for a realigning coalition. The risk is that his anti-establishment rhetoric alienates the party organization and its resources while energizing a coalition that may not be large enough to win a purple state. Jake Rourke's challenge is keeping Whitfield on message enough to win the general election without domesticating him so much that he loses the populist energy that makes him viable.

Both campaigns also face the polling challenges created by polarization: trying to understand an electorate where partisan nonresponse creates systematic biases, where self-described independents often behave like partisans, and where the nationalizing effect means that state-level polling numbers track national approval ratings more than they used to.

12.11 The Gap Between Map and Territory, Revisited

Let's return to where we started: the red-and-blue map. What does it actually tell us, and what does it hide?

What it accurately represents: the geographic distribution of partisan majorities. Most of the physical territory of the United States has a Republican plurality. Major cities and their immediate suburbs are Democratic strongholds. This geographic clustering is real, has grown over time, and has structural consequences for political representation.

What it obscures: the actual distribution of voters. Population-weighted maps tell a very different story — one where the Democratic-leaning areas are not islands but substantial population centers, where the moderate and competitive zones are larger than they appear, and where the appearance of overwhelming Republican geographic dominance is a visual artifact of low population density.

For the political analyst, the lesson is methodological: your unit of analysis matters enormously. If you analyze political geography at the county level, you will find a more polarized picture than if you analyze it at the individual level, because individual-level heterogeneity is suppressed when you aggregate to geographies. A county that votes 55-45 Republican looks solidly red on a map; at the individual level, it contains a substantial minority of Democrats who live their lives alongside Republican neighbors.

This is the gap between the map and the territory in its most concrete form. The map has consequences — it shapes how resources are allocated, how candidates campaign, what stories journalists tell. But it does not describe a politically uniform underlying reality. Purple is real, even when it doesn't look it.

Chapter Summary

The American partisan landscape has undergone a fundamental transformation since the mid-20th century. Partisan sorting — the alignment of ideology with party identity — has produced parties that are more ideologically coherent and more predictable than at any time in the survey era. Affective polarization — the growth of negative feelings toward the opposing party — has intensified partisan identity and made cross-partisan persuasion harder. The nationalization of elections has reduced the role of local candidate quality and increased the role of national partisan forces.

These changes are not just interesting academic findings — they have direct implications for how polls are conducted and interpreted, how campaigns are run, and how election outcomes are understood. Differential partisan nonresponse, social desirability bias, and herding are all exacerbated by polarization. The apparent persuadable electorate is smaller than self-reported independence suggests. The geographic sorting that produces the red-and-blue map conceals a more complex underlying electorate.

For the Garza-Whitfield race, polarization is both constraint and context: it shapes the strategic options available to both campaigns, complicates the polling environment, and determines the narrow slice of the electorate that will actually decide the outcome.

Chapter 13 turns to the demographic dimension of all this: who comprises the American electorate, how that composition is changing, and what the ongoing demographic transformation means for partisan coalitions — and for the analysts who try to understand them.

Key terms introduced: partisan sorting, ideological polarization, affective polarization, negative partisanship, nationalization of elections, the Big Sort, partisan nonresponse bias, social desirability bias, herding, partisan realignment, class realignment

Chapter 13 examines the demographic composition of the American electorate, the education realignment, and the dangers of "demographic destiny" thinking.

Extended Discussion: Measuring Polarization — A Methodological Deep Dive

The polarization literature is rich with competing empirical claims, and many of the disagreements trace to measurement differences rather than substantive disagreements. Before accepting any particular polarization finding, a careful analyst should understand what exactly is being measured.

Party Feeling Thermometers: The Standard Measure

The workhorse measure of affective polarization is the ANES party feeling thermometer: respondents are asked to rate each party (or its members) on a 0-100 scale, where 0 is very cold/unfavorable and 100 is very warm/favorable. The standard affective polarization measure is the difference between in-party and out-party ratings.

Several concerns attach to this measure:

Ceiling and floor effects: Strong partisans have historically rated their own party near 90 and the opposing party near 10-20. There is limited room for further movement at the extremes. If polarization appears to have plateaued, this may reflect measurement ceiling/floor effects rather than a genuine stabilization.

Scale interpretation shifts: The meaning of particular scale positions may change over time. In 1980, rating the opposing party at 35 may have reflected genuine mild unfavorability; in 2020, a 35 rating may reflect measured restraint from someone who actually feels quite hostile. The scale anchors are subjective, and their social meaning evolves.

Cross-study comparisons: The ANES feeling thermometer has been asked with somewhat different wording and context over the decades, making precise longitudinal comparisons difficult.

Alternative approaches to measuring affective polarization have been developed to address these concerns. Behavioral measures — willingness to hire, marry, or socialize with out-partisans — avoid the scale interpretation problem but introduce their own complications (stated intentions may not reflect behavior, and the settings are artificial). Implicit attitude measures (reaction time-based assessments of automatic associations) capture automatic affective responses but are hard to administer at scale in surveys.

Ideological Polarization: DWNOMINATE and Alternatives

For measuring elite polarization in Congress, political scientists rely primarily on DWNOMINATE scores — measures of legislator ideology derived from roll call voting patterns using an algorithm developed by Keith Poole and Howard Rosenthal. DWNOMINATE has the advantage of allowing comparisons across congresses going back to the founding, giving a long historical baseline.

The primary limitation: DWNOMINATE measures position on the floor of Congress, which reflects both true ideology and strategic voting considerations. A moderate Republican who votes with their party on procedural votes to maintain leadership goodwill will appear more conservative in DWNOMINATE than their true ideological position. As parties have become more cohesive, DWNOMINATE may increasingly reflect party loyalty rather than individual ideology.

For mass polarization, the ANES offers ideological self-placement (liberal-conservative scale), issue position batteries, and issue consistency measures. Each captures a different dimension of polarization. The absence of consensus on a single best measure partly explains why researchers reach different conclusions about how much mass polarization has occurred.

The Media Polarization Feedback Loop

One mechanism connecting elite polarization to mass affective polarization deserves extended attention: the role of partisan media.

The causal story runs roughly as follows: as elite politicians polarize, their rhetoric becomes more extreme and their portrayals of the opposing party become more negative. This content is amplified by partisan media outlets — cable news channels, online news aggregators, social media platforms — that are themselves economically incentivized to produce emotionally engaging content, and negative partisan content is among the most emotionally engaging.

Consumers of partisan media receive a diet of highly negative portrayals of the out-party that are not balanced by the routine cross-partisan contact of earlier eras (when geographic and social mixing was higher). Over time, their affect toward the out-party degrades — not because their policy positions have moved, but because the informational environment has made the out-party seem threatening, foolish, and malicious.

Research on the effects of partisan media is methodologically challenging because media consumption is not randomly assigned. Viewers who consume more partisan media may also hold more extreme prior opinions that led them to seek out that media. Experiments that randomly expose subjects to partisan media content find significant short-term effects on out-party affect; whether these short-term effects compound into long-run attitude change is harder to establish.

The social media dimension adds further complexity. Unlike cable news, which produces a broadcast product consumed by a defined audience, social media algorithms curate individualized content streams that may create the experience of extreme partisan conflict even for users who don't deliberately seek it out. A user who posts occasionally about politics may find their feed disproportionately populated with conflict and outrage, because conflict and outrage generate more engagement than balanced or moderate content. The platform's revenue optimization accidentally amplifies polarization signals.

The International Comparative Perspective on Polarization

Is the United States unusual in its level of polarization, or is affective polarization rising across all democracies? The comparative evidence paints a complex picture.

Studies examining affective polarization across democracies find that the United States has particularly high levels — among the highest of established democracies — but that polarization is rising in many countries simultaneously. This cross-national pattern suggests that some forces driving polarization are not uniquely American; they likely include the global rise of social media, the decline of class-based political identities, and the growth of cultural and identity-based political conflict that transcends national boundaries.

Countries with different electoral systems (proportional representation rather than first-past-the-post) and different party structures (multiparty systems rather than two-party systems) show somewhat lower affective polarization on average. This suggests that institutional structure matters: in a two-party system, political conflict is more zero-sum, identities are more binary, and the winner-takes-all dynamic makes partisan hostility more politically potent. Proportional systems allow for coalition governments that require cross-party cooperation, which may maintain more temperate affective relationships between parties.

🌍 Global Perspective: Polarization in Western Europe

Several Western European countries — most notably Poland, Hungary, Spain, and Italy — have experienced dramatic increases in polarization in recent years, sometimes producing democratic backsliding. In these cases, polarization has been more directly connected to ideological conflict (particularly over immigration, EU membership, and the cultural dimensions of globalization) than in the United States, where ideological and affective polarization have somewhat different tracks. The American case remains distinctive in the degree to which affective hostility has outpaced ideological distance, but the global pattern suggests that democratic institutions everywhere are being tested by partisan intensity.

Sorting, Polarization, and the Future of Competitive Elections

One of the underappreciated implications of partisan sorting and nationalization is what they mean for the long-run structure of electoral competition. In a sorted, nationalized environment, the number of genuinely competitive constituencies — at every level of government — shrinks.

Consider the trajectory: in the 1980s, there were roughly 100 competitive House districts, where the presidential vote in the previous election was within 5 points of the national average. By the 2020s, that number had fallen to roughly 30-35. The shrinkage reflects both partisan sorting (fewer moderates; more straight-ticket voters) and geographic sorting (voters clustering into more homogeneous communities).

The consequences for democratic representation are profound. Elected officials from non-competitive districts face their most consequential elections in primaries, where the electorate is more partisan and more ideologically homogeneous. This creates incentive structures that reward ideological purity over cross-partisan governance. The institutional mechanics that once encouraged moderate representation — competitive general elections in which the median voter exerted discipline — have weakened.

For the competitive constituencies that remain — like our Sun Belt state in the Garza-Whitfield race — the national stakes have paradoxically increased. With fewer competitive states and districts, each one carries more electoral weight. Campaign resources, candidate quality, and analytical sophistication are increasingly concentrated in a small number of arenas where the national balance of power is at stake.

This concentration creates a feedback loop: competitive constituencies receive more investment, more attention, and better representation, while non-competitive constituencies — regardless of their objective policy needs — receive less. The political economy of the competitive era increasingly functions as a winner-take-all tournament, with a few arenas absorbing most of the resources and attention of both parties.

Jake Rourke Reflects: Polarization From the Practitioner's View

Eight weeks before election day, Jake Rourke allowed himself a rare moment of reflection in a conversation with an old colleague from his first Senate campaign — back in 2002, when things were different.

"Back then," Jake said, "I actually had to think about message. What do we say to soft Republicans who might peel off? What do we say to conservative Democrats in the rural counties? There were real people in those boxes who were movable."

"Now," he continued, "the game is mostly about juice. Whose base is more energized. Whose people actually turn out. Whitfield doesn't need to persuade anyone in the traditional sense — he needs to make his people so angry about what the other side represents that they get off the couch and vote."

His colleague asked if he thought that was sustainable — a politics based entirely on negative mobilization, with very little persuasion at the margins.

Jake was quiet for a moment. "I don't think it's sustainable long-term. You can't govern a country that way — you need to build coalitions, make deals, do the actual work. But in any individual election cycle, it's the rational strategy. Nobody wants to be the one campaign that plays the chess game while everyone else plays checkers."

This practitioner's view captures something important about the relationship between polarization research and campaign strategy. The academic literature has documented the rise of negative partisanship and affective polarization. Campaigns have, rationally, adapted to that environment by increasingly targeting negative mobilization over positive persuasion. But those adaptive strategies also reinforce the polarization they respond to — by further degrading the stock of cross-partisan goodwill and by giving voters fewer reasons to think of the opposing party as legitimate rather than threatening.

Whether and how this feedback loop can be interrupted is one of the central questions of contemporary democratic theory. It is also, more practically, the challenge that any reform effort — from ranked-choice voting to open primaries to campaign finance reform — must address if it wants to change the strategic incentives that produce the behavior we observe.

Analytical Applications: Polarization Indicators for Campaign Monitoring

For the working political analyst, several specific polarization indicators are worth tracking throughout a campaign cycle:

Party ID gap trends: Track the gap between the percentage of survey respondents identifying as Democrat minus those identifying as Republican over time. Widening toward your party suggests improving enthusiasm; narrowing suggests warning signs.

Enthusiasm differential: Compare the percentage of each party's identifiers who say they are "extremely" or "very" interested in the upcoming election. When your party's enthusiasm significantly leads the other party's, mobilization pays off; when it lags, you face a turnout environment headwind.

Generic congressional ballot: In environments with strong nationalization, the generic congressional ballot (do you plan to vote for the Democratic or Republican candidate for Congress in your district?) is a reasonable proxy for the partisan environment's overall lean. It translates roughly to Senate and governor's races in competitive states.

Presidential approval tracking: In a nationalized environment, presidential approval is a strong predictor of the partisan environment. A Democratic president at 52% approval creates a favorable environment for Democratic Senate candidates; at 45%, it's a headwind.

Cross-party evaluations: Track how each party's identifiers evaluate the opposing party's candidates — not just topline horse-race numbers. When these cross-evaluations soften (out-partisans becoming slightly less hostile to the opposing candidate), persuasion may be working. When they harden, the election is shifting toward pure base mobilization.

Nadia tracks all of these weekly, feeding them into a dashboard that she reviews with Garza's campaign manager every Monday morning. The indicators don't tell her what the outcome will be — no set of leading indicators does that. They tell her what type of environment the campaign is operating in and which strategic adaptations that environment calls for.

Extended Analysis: Is American Democracy Distinctively Polarized?

To place the American case in perspective, it is worth asking whether the United States is an outlier in global terms or whether the polarization patterns described in this chapter reflect broader trends visible across democracies.

The short answer: both. The United States is at the high end of affective polarization among established democracies, but it is not alone; many democracies are experiencing rising partisan hostility. What makes the American case distinctive is the combination of factors that amplify polarization's political consequences.

Two-Party Systems and Polarization Dynamics

Political scientists have long noted that two-party systems produce different polarization dynamics than multiparty systems. In a two-party system, the parties are the only available identities for partisan attachment. When citizens categorize themselves into one of two groups, social identity theory predicts strong in-group favoritism and out-group hostility — the binary structure maximizes the identity-threat dimension of partisan conflict.

In multiparty systems, citizens distribute across more parties. When a right-of-center party governs, it typically does so in coalition with one or more centrist parties — requiring compromise, cross-party negotiation, and the regular visibility of cooperative inter-party relationships. This institutional structure may dampen the zero-sum character of partisan conflict even when ideological differences are significant.

The United States' two-party system — reinforced by plurality electoral rules, winner-take-all elections, and two-party ballot access advantages — creates a structural predisposition toward binary partisan conflict that multiparty systems may partially avoid.

Electoral Institutions and the Polarization-Accountability Tradeoff

When political scientists and reformers discuss potential remedies for polarization, electoral institutions are frequently raised: ranked-choice voting, open primaries, proportional representation, independent redistricting, and other changes are advocated as polarization-reducers.

The empirical evidence on these reforms is mixed but not discouraging. Alaska's adoption of ranked-choice voting and a top-four open primary in 2020 produced election outcomes that were less partisan-extreme than comparable states using closed primaries. California's top-two primary produced some measurable moderation in the candidates that survived to general elections in safe districts, though the effect was smaller than advocates hoped.

The larger point for the analyst is that polarization is not simply a feature of public opinion that campaigns must accept as fixed. It is partly a product of institutional structures that shape incentives for candidates, parties, media, and voters. Understanding those structural drivers is essential for evaluating whether — and how — the polarized environment in which campaigns operate might change over time.

For now, in the election cycle of the Garza-Whitfield race, the structural environment is given: the institutions are what they are, the media ecosystem is what it is, and the partisan identities of most voters were formed in a world that has been sorted for decades. Campaigns adapt to that environment. But the analyst's role includes understanding why that environment exists — not just mapping its contours but tracing its causes — because the causes determine which changes, if any, are feasible.

The Analyst's Responsibility in a Polarized Environment

We close with a question that sits at the intersection of professional practice and civic responsibility: what obligations does a political analyst have in a polarized environment?

On one view, the analyst's job is purely technical: describe the electorate accurately, forecast outcomes reliably, help campaigns allocate resources efficiently. Questions about the health of democracy are for others — politicians, journalists, citizens, reformers.

On another view, the analyst occupies a distinctive epistemic position — they know more about how partisan psychology works, how polarization operates, and how campaigns exploit tribal instincts than most participants in democratic politics. With that knowledge comes a form of responsibility: to use it honestly, to be transparent about its limits, to resist using analytical sophistication to amplify rather than understand partisan manipulation.

The tension between these views doesn't resolve cleanly. But the most thoughtful practitioners — Vivian Park at Meridian, Adaeze Nwosu at ODA, and, on her best days, Nadia Osei at the Garza campaign — navigate it by maintaining a distinction between their professional obligations to their clients and employers, and their broader commitments to truthful representation of what the data shows.

A pollster who suppresses a finding because a client doesn't like it, or who adjusts a topline number to match a desired narrative, has crossed a line that cannot be uncrossed. An analytics director who tells a campaign only what it wants to hear, rather than what the data shows, is ultimately serving neither the campaign nor the democratic process. And a data journalist who presents polarization figures without context — in ways that make the situation seem more hopeless or more amenable to simple solutions than it is — is using analytical authority to shape rather than illuminate public understanding.

These are professional ethics questions, and they are particularly pointed in a polarized environment, precisely because polarization makes truthfulness more costly and selective presentation more tempting. The remaining chapters of this book return to these questions in various forms. Chapter 38 addresses them directly. For now, holding them clearly in view as we examine polarization's specific manifestations is sufficient — and necessary.

A polarized electorate is not just a data pattern to model. It is a fact about the civic environment in which analysis happens, with consequences for what can be honestly said and what obligations attach to saying it.

Extended Methodological Note: Tracking Polarization Over Time — Best Practices

For the analyst who wants to rigorously track polarization trends in a specific state or constituency, this section offers practical guidance on data sources, measurement choices, and common errors.

Data Sources

ANES Time Series: The gold standard for measuring mass partisanship, ideology, and affective polarization over time. The feeling thermometer batteries and party identification measures have been consistently asked since 1952 (party ID) and the 1960s-1970s (feeling thermometers). Available free at electionstudies.org.

Pew Research Political Polarization Studies: Pew has fielded large-sample surveys specifically designed to measure partisan divides in opinion on a wide range of issues. Their 2014 and 2019 Political Polarization reports are particularly detailed. Available at pewresearch.org.

DWNOMINATE Congressional Roll Call Data: For measuring elite polarization in Congress, Poole and Rosenthal's DWNOMINATE scores provide the most comprehensive longitudinal dataset. Available through the VoteView database at voteview.com.

State-Level CCES: The Cooperative Congressional Election Study (CCES) surveys 50,000+ Americans per election cycle, with state-level samples large enough for meaningful sub-national analysis. Available through the Harvard Dataverse.

Catalist and Voter File Analytics: For campaign-level analysis, commercial voter file providers like Catalist maintain historical data on voter file partisanship and turnout patterns that allow reconstruction of multi-cycle trends in specific geographies.

Common Measurement Errors

Conflating sorting with polarization in cross-tabs: If you observe that Democrats and Republicans hold increasingly different positions on immigration, you cannot conclude that either group has moved toward the extreme. It may simply be that people who held liberal positions sorted into the Democratic Party. Always check whether the positions of "liberals" and "conservatives" have changed, not just the positions of "Democrats" and "Republicans."

Using party feeling thermometers without in-party controls: Affective polarization is conventionally measured as the in-party thermometer minus the out-party thermometer. But if in-party ratings have remained stable (or even declined slightly) while out-party ratings have fallen, the total gap can widen even as people like their own party less. Reporting only the gap obscures this nuance. Report both components.

Treating 2016 as the origin point: Much political commentary treats polarization as something that happened after 2016 — a consequence of the Trump era. In fact, both ideological sorting and affective polarization were well underway by 2000 and accelerating through 2008 and 2012. Trump may have accelerated certain dimensions of polarization, but presenting 2016 as year zero produces a badly distorted historical picture.

Ignoring survey mode effects: Online surveys and phone surveys sometimes produce different estimates of partisan feeling. If you switch survey mode over time, apparent changes in polarization may reflect mode effects rather than real opinion change. Maintain methodological consistency when tracking trends.

A Practical Polarization Dashboard for Campaign Analysts

For a campaign operating in the Garza-Whitfield environment, a weekly polarization monitoring dashboard might include the following indicators:

1. Statewide generic ballot: D% minus R% among likely voters. The trend line over the campaign period tells you whether the partisan environment is stable, moving Democratic, or moving Republican.

2. Enthusiasm ratio: Among people who "definitely plan to vote," the ratio of Democrats to Republicans. A ratio above 1.1:1 indicates Democratic enthusiasm advantage; below 0.9:1 indicates Republican advantage. Track weekly.

3. Out-party unfavorability: Percentage of each party's identifiers who rate the opposing party's candidate as "very unfavorable." Rising numbers indicate intensifying negative partisanship — which typically benefits mobilization but may complicate persuasion.

4. Independent breakdown: Among self-described independents, the partisan lean of their vote intention. When "independents" are voting 60-40 for one party, the electorate is more polarized than the self-ID data suggests. This is the best single diagnostic for the size of the genuine persuadable universe.

5. Presidential job approval by party: The approval gap between Democrats and Republicans evaluating presidential performance. A wide gap (Democrats 85% approve, Republicans 5% approve) indicates the polarized environment is high; a narrower gap suggests either a rally effect or genuine bipartisan evaluation.

These five indicators, tracked weekly, provide the analyst with a real-time reading of the partisan temperature that contextualizes all other analytical work. They don't replace deeper analysis — but they ensure that deeper analysis is conducted in accurate awareness of the competitive environment.