Introduction & Methodological Framework

What the executive-branch flip actually shows

The opening paragraph is doing two jobs at once. The smaller job is documenting a fact: an American's confidence in their own government's executive branch is largely a function of whose team is in charge of it. The larger job is illustrating the kind of question this book is built to answer. Headlines about polarization and "the great divide" appear roughly every news cycle. Most of them rest on cross-sectional data — a single poll, a single year, a single moment — and most of them have a half-life of about a week. Knowing whether a given moment is part of a trend, the end of one, or a temporary excursion requires a long enough series to tell those things apart.

The General Social Survey, conducted by NORC at the University of Chicago since 1972, is the longest such series in the United States. It is also unusually broad: the same instrument that asks about confidence in the executive branch also asks about happiness, social trust, abortion, gun ownership, racial attitudes, religious practice, and dozens of other things. Because the GSS holds question wording and sample design constant across waves, it lets us trace not just one issue's trajectory but the correlations between issues — whose attitudes hang together, and how that bundling has changed.

A few things in Figure 1 are worth slowing down on.

The lines genuinely cross. This is not a chart where one party is consistently more trusting and the other consistently less; the relative positions invert with each administration. That is the partisan-perception story in its simplest form. It also means that the "trust in government has collapsed" framing common in op-eds is, on this measure, a half-truth. Trust in the executive branch has not so much collapsed as become a function of the partisan match between the respondent and the office-holder.

The second feature is the floor. Look at the loser's line in any given year — the party that does not control the office. In the 1970s and 1980s, the out-party's confidence in the executive branch typically sat between 10% and 20%. By the 2020s, it routinely sits below 10%, and in 2024 it falls to 5%. Whatever else is happening, the partisan losers are losing harder.

The third feature is harder to see in the chart and easier to see in the table that backs it: the winner's confidence has not risen by the same amount that the loser's has fallen. The Obama-era Democratic peak (25% in 2010) is not far from the Reagan-era Republican peak (33% in 1982). The Biden-era Democratic peak (21%) is, if anything, slightly lower. What has grown is not so much partisan enthusiasm as partisan repulsion.

The pattern is not unique to the executive branch

If you look only at the executive branch, the partisan-flip story is so clean that you might mistake it for the whole story. It is not. Trust in Congress — a body whose composition does change every two years, but whose institutional shape is fixed — has not flipped at all. It has simply collapsed, on both sides.

Read together, Figures 1 and 2 are doing different work. The executive-branch chart is about who trusts. The Congress chart is about whether anyone does. One of the running themes of this book is that the standard polarization narrative — Americans are sorting into two opposed camps — is the right story for some questions and the wrong story for others. For executive trust, the camps are real, the gap is large, and the gap is widening. For trust in Congress, the camps barely matter, because nobody is in either of them. A book that cannot distinguish between those two patterns will systematically misread which institutions are in trouble and what kind of trouble they are in.

Sorting on culture: the abortion case

The other story this book has to tell is the one in its title: that over the same five decades, Americans have sorted — meaning that issues that used to be only loosely correlated with party are now tightly correlated with it. The cleanest single illustration of this in the GSS is the question about whether a woman should be able to obtain a legal abortion "for any reason."

In the late 1970s, this question genuinely crossed party lines. In 1977, 42% of Republicans answered yes; only 35% of Democrats did. (Those numbers are not transposed — Republicans really were the slightly more pro-choice party, on this question, in this year.) The cross-pressures of that era are gone. By 2024, 81% of Democrats answered yes, against 33% of Republicans — a 47-point partisan gap on a question where, fifty years earlier, the parties had been within a few points of each other and pointing in the opposite direction.

What Figure 3 makes clear, and what most coverage of the "abortion divide" misses, is that the divergence is largely one-sided. Republican support for legal abortion in 2024 (33%) is below Republican support in 1977 (42%) — a 9-point drift down. The Democratic line, in contrast, more than doubles over the same period, from 35% to 81%. The polarization on this question is not a story of two camps moving symmetrically in opposite directions; it is mostly a story of one camp moving and the other holding roughly steady while the political coalitions reshuffled around them. The chapter on cultural sorting (Chapter 4.1) returns to this question in detail, including breakdowns by religion, age, and education.

A teaser: well-being follows ideology

The last thing this opening chapter wants to do is preview that the book's reach extends beyond political attitudes. The GSS asks how happy you are. The answer turns out to depend, in a reasonably orderly way, on what you believe.

The full version of this story is in Chapter 7.2. The short version, for now, is that the Conservative-minus-Liberal happiness gap was 9 points in 1974, sat between 5 and 10 points for forty years, then widened sharply during the 2010s — driven primarily by the Liberal line falling. By 2024 the gap had narrowed somewhat, but neither side has returned to the 1970s levels. Whether American well-being has a partisan component, or whether partisan well-being has an American component, is a question the data do not fully answer; but the data are clear that the question is real.

What the rest of this book does

Each chapter ahead picks a domain — institutional trust, government spending, civil liberties, race, religion, gender, education, geography — and asks the same set of questions of it.

1. What is the long-run trajectory of the topline measure? (Has the average American moved?)
2. How does that trajectory differ by demographic? (Who has moved and who has not?)
3. Has the partisan gap on this issue grown, shrunk, or held steady? (Is this a sorting story, a collapse story, or neither?)
4. Where the answer is interesting, what is the closest thing the GSS can say about why?

The "what" of polarization is settled enough that there is little new to be said about it. The "where" and the "how much" are not. American polarization is uneven across topics, uneven across demographics, and uneven across decades; some divides that loom large in the discourse are, on the data, narrower than they look, and some that loom small are wider. Sorting these out one figure at a time is what the rest of this book does.

A short note on the data

A complete methodological appendix lives at the back of the book; this section is the minimum a reader needs to be able to interpret the figures.

Population. The GSS samples the noninstitutionalized adult population of the United States. Throughout this book, "Americans" means "GSS respondents weighted to that population."

Weights and sampling design. All point estimates use the GSS sampling weight (wtssps); all confidence intervals account for the survey's stratification (vstrat) and clustering (vpsu) via Taylor series linearization. Treating GSS data as a simple random sample — a common shortcut in journalism that uses the data — produces standard errors that are too small by 30 to 50 percent.

The 2021 mode shift. GSS data collection moved from in-person to web-based in 2021 in response to the pandemic. On objective items (year of birth, region of residence) the mode shift is irrelevant. On subjective items (happiness, trust, satisfaction) the same person interviewed by web tends to give somewhat lower readings than the same person interviewed in person. We treat 2021 as comparable to 2022 and 2024 (all web) but not directly comparable to 2018 and earlier (all in-person). Where a particular finding hinges on this, the relevant chapter says so.

Substantive vs. statistical significance. The GSS is large enough that small effects are routinely statistically significant. We follow the APA's preferred convention of leading with effect sizes — point differences, percentage-point gaps, and trends — and mention p-values only when a finding's reliability is genuinely in question. A 5-point shift in 'very happy' is meaningful even if not statistically significant in a single year; a 0.5-point shift across a 30,000-respondent pooled estimate is statistically significant but substantively trivial.

The figures throughout the book follow a small set of conventions: blue for Democrats and Liberals (red for Republicans and Conservatives), a faint background banding by presidential era, a hover tooltip with sample size, and confidence intervals shown as lighter ribbons or error bars. The next page starts with the institutions whose collapse Figure 2 hinted at — Congress, the press, science, the church, and the others — and asks who lost faith first.