Money pulls the parties along different axes

Within-party income slope — the shift in each index from the bottom to the top of the income distribution, net of education, age, race, region, and year — in standard deviations. Bars are weighted WLS estimates with HC1 robust 95% intervals. GSS.

Democrats Independents Republicans

Takeaway: on the economic index (left) the Republican bar towers over the others — income makes Republicans much more economically conservative. On the social index (right) the Democratic bar is tallest and the Republican bar nearly vanishes — income makes Democrats more permissive but does little to Republicans.

Slopes are the marginal effect of weighted income rank (0 = poorest to 1 = richest) within each party, from WLS with HC1 robust standard errors and survey-year fixed effects. Higher economic = more conservative; higher social = more permissive. Source: GSS, weighted by weight_use; n ≈ 62,000.

Within each party, the pew divides more than the paycheck

Weighted mean of the social index for the more-observant and less-observant half of each party. Up = more permissive. The connecting bar is the religious gap. GSS.

Takeaway: in both parties the more-observant half sits far below the less-observant half on the permissiveness scale — a gap of roughly six-tenths of a standard deviation, opening the same way on each side of the aisle.

More- vs less-observant split at the weighted median of church attendance within each party. Source: GSS, weighted by weight_use. In ANES the same split yields religious gaps of +0.47 (Democrats) and +0.45 (Republicans).
Economic
Social

 

The income gradient, by party

The question’s weighted mean across income thirds, one line per party. Up = the conservative or traditional side (oriented to match the index).

The religious split, by party

The question’s weighted mean for the more- and less-observant half of each party. The connecting bar is the religious gap.

Each item is weighted-z-scored and oriented by its own correlation with party. GSS, pooled 1972–2024, weighted by weight_use; groups below 150 (gradient) or 100 (religion split) suppressed. For economic items the religious split is typically small — religion governs the social domain, not the economic one.