Start here

In American surveys, a person's party says more about the ideological label they choose than it did in the 1970s. It still leaves much unknown about their positions on particular issues, how they feel about the other party, and the kind of place where they live. The Public Mind is about those relationships: how closely party, ideological label, issue positions and the political map line up, and where they stay distinct.

The book follows party labels into beliefs, social groups, issues and attitudes toward democracy, then turns to election returns, political donors and the people who live in different places. Its evidence is survey data, county and precinct returns and campaign finance records. The chapters ask where these records agree and where the answer changes with who is counted, what is compared, the period or the geographic unit. Each chapter asks one question, answers it with figures, and ends with what the data cannot show.

41 of 42 chapters are written, all of them drafts that are still being checked. The other 1 is planned: it shows the question it will ask and the data it will use, and states no result. Further questions are on the research agenda.

Three ways in

1. Read in orderBegin with the introduction, Party, belief and the political map, and follow the book part by part.
2. Start with the map5 chapters that read the vote through the county map.
3. Take the grand tour8 chapters, one from each part.

All chapters lists every chapter with its data and status. Explorers are interactive tools that sit beside the parts, and Reference holds the data sources, glossary, figure gallery, methods and the research agenda.

The short way through the argument is the core reading path, from party and ideology through left and right across countries to the county map and the people who live in it.

A figure from the book

Counties can move with the nation from one election to the next and still sit far from it. The map below, from Chapter 6.2, Distance from the national vote, 1868 to 2024, shows which counties were far from the national margin, and its slider compares 2024 with 1976.

The three volumes

Volume 1

Labels, sorting and partisan feeling

Party identification and ideological labels drew closer together in American surveys. The volume asks what the labels contain, in the United States and about 100 countries, how the alignment changed across birth cohorts, and how it differs from partisans' feelings toward the other party.

1. Part I

   Labels and beliefs

   Party identification and ideological labels became more closely aligned, yet the labels have a loose fit with what people want and different contents in different countries.

   5 written, 0 planned · World values explorer

2. Part II

   Change and partisan feeling

   The alignment grew within birth cohorts, and warmth toward one's own party moved little while warmth toward the other party fell, in every group examined.

   3 written, 0 planned

Volume 2

Coalitions and political differences

Participation, group coalitions, issues and attitudes toward institutions and democracy, 1948 to 2024. Party gaps on these measures differ in size and timing, and some reverse with control of the presidency, so no single trend describes them.

1. Part III

   Participation and coalitions

   Who votes, and who votes for each party, changed along education, income, race, gender, age and religious attendance, and attendance relates to the vote by different amounts inside different religious traditions.

   8 written, 1 planned

2. Part IV

   Issues move on different schedules

   Each issue, the racial-attitude items among them, has its own partisan gap, its own timing and its own demographic profile.

   8 written, 0 planned · Issues explorer

3. Part V

   Institutions and democratic attitudes

   Party gaps in confidence widened institution by institution, approval of strong leaders rose across countries while stated support for democracy changed less, and justifying political violence shows no detectable party gap.

   4 written, 0 planned

Volume 3

Votes, giving and places

Election returns, donor records and county data count different populations at different scales. The volume follows the vote from counties to precincts and down the ballot, compares donors with voters, and returns from places to the people who live in them.

1. Part VI

   Election results at different scales

   County swings moved more nearly in step in 2020 to 2024 than in any pair of presidential elections since 1868, while a rising share of votes is cast in counties far from the national result, and a county margin is an average over precincts that are often far more lopsided than the county.

   4 written, 0 planned · Elections, 1976 to 2024 explorer

2. Part VII

   Giving

   Campaign donors are concentrated in a few hundred college-heavy counties, and their pool sorted before the electorate did.

   2 written, 0 planned

3. Part VIII

   Places and the people in them

   Education, income, ethnicity and rurality each relate to the county vote differently from what the raw correlation suggests, and within a party, where someone lives says far less about their views than which party they belong to.

   7 written, 0 planned · Neighborhoods explorer

Rules of evidence

Survey estimates are weighted, and each one carries a confidence interval and the number of respondents behind it. Hover over or focus any number in the text to see its estimand, interval, sample size, weights and source. No number in the prose is typed by hand. Each is printed from a data file that the analysis writes, so a number in the text cannot differ from the number in the figure through a transcription error. The file does not check that a sentence names the right population, denominator or interpretation for that number, and those are reviewed separately. A figure's "Show data" table lists the exact values behind every mark.

The language is descriptive. A chapter says that one group is more likely than another, or that a gap widened, and does not say that one thing caused another unless the design supports it.

Every chapter is labeled with its grain, which sets what a claim can be about: individuals, places, countries or officeholders. Maps describe the composition of places, and a county that votes one way says nothing about any one voter in it. Donor records describe the people who gave, who are not a sample of all Americans. Two surveys changed how they collect answers during the years the book covers, and figures that cross such a break mark it. The methods page gives the full grain ladder, the survey mode breaks and the rules for small cells.

How to read a figure

Each figure has a title that states its finding, a subtitle that says what is plotted, and a source line that gives the dataset, the years and whether the estimates are weighted. Lines and bars are labeled directly. Shaded bands are confidence intervals. Color is never the only cue: the same information appears in labels, in the hover text and in the data table.