Two outliers, opposite corners

Every party × income-third × college cell, placed by its weighted mean on the economic index (horizontal, right = more economically conservative) and the social index (vertical, up = more socially permissive). The origin is the national center; larger rings mark the two cells furthest from it. GSS, pooled 1972–2024.

Democrats (incl. leaners) Republicans (incl. leaners) Independents ◯ ring = college

Takeaway: the cloud is not a diagonal line. The affluent college Democrat floats to the top-left (socially permissive, economically moderate); the affluent college Republican sits at the right (economically conservative, socially middling). They are unlike each other and unlike everyone else — on different axes.

Indices are weighted z-scored composites (higher economic = more conservative; higher social = more permissive); each item is oriented by its own weighted correlation with party. Cells with fewer than 150 respondents or a Kish effective N below 100 are suppressed. Source: GSS, weighted by weight_use.

ANES finds the same two corners

The identical construction in the ANES Cumulative File. The outliers are again the affluent college-educated Democrat (top) and the affluent college-educated Republican (right).

Takeaway: two independent surveys, built from different question wordings and weighted differently, place the same two cells in the same opposite corners.

In ANES the affluent college Democrat reaches +0.82 sd on the social index and the affluent college Republican +0.71 sd on the economic index. Source: ANES Cumulative File, weighted by weight.
Economic
Social

 

 

Where each group stands on this question

Weighted mean for every party × income-third × college cell, in standard deviations (right = the conservative or traditional side, oriented to match the index).

Dem Ind Rep ◯ ring = college

Does income move the parties the same way?

The question’s weighted mean across income thirds, one line per party. Diverging lines mean income sorts the parties differently on this item.

Each item is weighted-z-scored and oriented by its own correlation with party, so the scale is comparable across questions and to the index. GSS, pooled 1972–2024, weighted by weight_use; cells below 150 respondents suppressed.