The Great Sorting

Ch. 14

The Multidimensional Divide

The standard model of American political ideology is one-dimensional: there is a Liberal–Moderate–Conservative axis, and voters' positions on it predict their positions on every major issue. The model has been the workhorse of academic political science since the 1950s, and it explains a substantial fraction of what survey research observes. But this book has documented several findings where the single-dimension model breaks down — where two voters who share an ideological label disagree systematically on a specific issue, or where a single voter's position on one axis (cultural conservatism, say) does not predict their position on another (economic conservatism). This chapter is about the cases where the standard model fails and about what those failures suggest about the actual shape of contemporary American political space.

2+axes
The minimum number of dimensions required to describe the contemporary American political space. The standard left-right axis explains most variance on most issues, but several major issues require an additional dimension to describe — most clearly, an authority-vs-autonomy axis that cuts across the standard economic-conservatism axis, and an in-group-vs-out-group axis that overlaps with race but is not reducible to it.

Where the one-dimensional model works

A point worth making clearly: the one-dimensional left-right model works very well for most of what this book has documented. The chapter-by-chapter findings, when summed, show that knowing a respondent's ideological self-identification (or their party identification, which §12 documented has nearly fused with ideology) is enough to predict their position on most issues most of the time.

On economic-spending questions, the left-right axis predicts attitudes well. On cultural-attitudinal questions about race, abortion, LGBTQ rights, and gender, the same axis predicts attitudes well. On institutional-confidence questions about science, medicine, and the press, the same axis predicts attitudes well. The contemporary American Liberal and Conservative coalitions hold systematically different positions across the GSS battery, and those positions cluster on the single dimension.

The cases where the one-dimensional model fails are interesting precisely because they are exceptions to a generally strong pattern.

Failure case #1: the authority-vs-autonomy axis

The clearest failure of the one-dimensional model is on questions about order, authority, and force. Two voters who agree on environmental regulation can disagree sharply on police use of force; two voters who agree on abortion access can disagree sharply on civil-liberties tolerance for unpopular speech.

The clearest example in this book: the §5 chapter on police authority showed that the contemporary partisan gap on force in some situations (polhitok) is 20 points, with Republicans more permissive. But within the Liberal coalition, sub-coalitions disagree — younger Liberals are sharply less tolerant of police force than older Liberals are, and libertarian Liberals (a small but growing subset of the Democratic coalition) hold positions on police force that look more Republican than Democratic. The standard left-right axis does not capture this within-Liberal disagreement; an additional authority-vs-autonomy axis does.

The §4.2 chapter on civil liberties produced the parallel finding: Democrats are 14 points less tolerant of racist speech than Republicans, but 13 points more tolerant of anti-American Muslim cleric speech. A single ideological axis cannot put one party in both positions simultaneously. A second axis — "tolerance of unpopular speech threatening to my coalition" — predicts the pattern.

The authority-vs-autonomy axis is largely orthogonal to the standard left-right axis. It produces:

  • Liberal-Libertarians (high autonomy on cultural and economic issues)
  • Liberal-Authoritarians (high autonomy on cultural issues, willing to use government force for economic and social ends)
  • Conservative-Libertarians (high autonomy on economic issues, traditional on cultural issues)
  • Conservative-Authoritarians (high authority on both axes)

In the contemporary American electorate, the modal Democrat is a Liberal-Libertarian on economic issues but increasingly Authoritarian on cultural enforcement (the contemporary "cancel culture" debates). The modal Republican is a Conservative-Authoritarian on most issues with Libertarian leanings on specific economic items (taxes, regulation). The cross-pressured cells — Liberal-Authoritarian Democrats and Conservative-Libertarian Republicans — are the political space where the one-dimensional model fails.

Failure case #2: the in-group-vs-out-group axis

The §10 chapter on White voters documented that the contemporary partisan map runs heavily on racial composition. Black voters of any ideology vote substantially more Democratic than White voters of the same ideology. The standard left-right axis predicts ideology, but ideology does not predict party identification within racial groups in the same way it does across them. The race-of-respondent variable is a meaningful additional axis.

The §4.4 chapter on racial attitudes showed the same axis from another angle. White Republicans and White Democrats disagree by 48 points on the discrimination-as-explanation question. Black Republicans and Black Democrats disagree by much less — the racial dimension dominates the partisan one within the Black population. The one-dimensional ideological model treats Black Republicans and White Republicans as belonging to the same coalition; the actual data shows them holding substantially different positions on race-related questions.

The in-group-vs-out-group axis is a feature of the composition of the parties more than of the positions the parties hold. It manifests as: voters whose own group is the majority of one party hold positions that align with that party's platform more strongly than voters whose group is a minority of either party. The contemporary Democratic Party is more racially diverse than the Republican Party; its coalition therefore contains more cross-pressured voters on race-related questions.

Failure case #3: the religiosity axis

A finding that recurs across multiple chapters in this book: the church-attendance axis predicts attitudes on certain questions better than party identification does. The clearest cases:

  • Abortion (§4.1): the partisan gap on legal-abortion-for-any-reason is ~40 points; the attendance gap is roughly the same.
  • Homosexuality always wrong (§4.3): the partisan gap is 33 points; the attendance gap is 49 points.
  • Premarital sex always wrong (§4.3): the partisan gap is 20 points; the attendance gap is 39 points.

On these questions, a respondent's church attendance is a stronger predictor of their position than their party identification. The contemporary American political space contains a religiosity axis that overlaps with but is not reducible to the standard left-right axis. The "secular Democrat" and the "secular Republican" are closer to each other on sexual-morality questions than either is to a weekly-attending member of their own party.

The religiosity axis has compressed over the past fifty years — fewer Americans attend church weekly, and the secular share of both coalitions has grown. But where it operates, it operates strongly.

What the multidimensional model predicts

A model of the contemporary American political space with at least three dimensions — left-right economics, authority-autonomy, and religiosity — would predict several patterns the one-dimensional model does not:

  1. The contemporary Liberal coalition is internally fractured on questions of authority. Younger Liberals and older Liberals disagree on enforcement-of-cultural-norms questions in ways that the standard one-dimensional model cannot capture.

  2. The Conservative coalition is internally fractured on questions of religiosity. Secular Republicans (a growing share of the contemporary GOP) and weekly-attending Republicans disagree on sexual-morality questions in ways the one-dimensional model misses.

  3. The "post-liberal" reorganization sometimes discussed in contemporary political journalism is not a third party emerging in the standard left-right space; it is a coalition emerging in the authority-religiosity sub-space that the one-dimensional model treats as part of the standard right. The Sohrab-Ahmari / Adrian-Vermeule "post-liberal" intellectual movement, the contemporary Republican social-conservative wing, and the post-Trump nationalist tendency all live in the authority-religiosity sub-space.

  4. The "online progressive" coalition does not have a clean position on the standard left-right axis. It is highly liberal on cultural questions, mixed on economic questions, and high-authority on enforcement-of-cultural-norms questions. It lives in a different sub-space than the New Deal Liberal coalition does.

The GSS's limits as a multidimensional instrument

The GSS measures left-right ideology with a single seven-point self-identification scale. It does not directly measure the authority-autonomy axis, the religiosity axis (beyond church attendance), or the in-group-vs-out-group axis as separate dimensions. The book has had to infer the existence of these additional axes from the GSS's specific-issue questions, which is a less clean methodology than directly measuring them would be.

The next generation of political-attitude surveys — particularly the work coming out of Pew Research's typology studies and the More in Common's "Hidden Tribes" research — has been more deliberate about multi-dimensional measurement. The findings from those instruments are broadly consistent with the inferences this book has drawn from the GSS, but they offer cleaner direct measurement of the dimensions the GSS only indirectly captures.

What this chapter changes

Two observations to take into the final chapter.

First, the contemporary partisan sorting (§12) is mostly but not entirely a sorting on a single axis. The cross-pressured voters who used to share both parties' platforms are gone; the contemporary Democratic and Republican coalitions hold internally consistent positions across most issues. But on specific issues — police force, civil-liberties tolerance, sexual morality — within-coalition disagreements remain that the standard model cannot describe.

Second, the within-coalition disagreements may be where the next decade of American political change happens. The contemporary Democratic coalition contains a Liberal-Authoritarian sub-coalition (the "online progressive" tendency) that is in tension with the Liberal-Libertarian sub-coalition (the older "ACLU Democrat" tradition). The contemporary Republican coalition contains a Christian-Conservative sub-coalition (weekly church-attending traditional moralists) that is in tension with the secular-Republican sub-coalition (the post-Trump nationalist tendency). Whichever sub-coalition wins within each party will shape the next generation of partisan platforms more than any single between-party shift will.

The final chapter — §15 — closes the book by summarizing the major findings of all fourteen chapters and considering what they imply for the next decade.