The Great Sorting · Religion
The religious fault line in American voting did not fade — it rotated. The mid-century divide between Catholic Democrats and Protestant Republicans has given way to one between the devout and the secular, a line that now runs straight through every pew and parish.
In 1960, a Catholic and a Protestant who attended church with identical fervor would still have voted very differently: the Catholic far more Democratic, the Protestant far more Republican. Which church you belonged to was the thing that moved your vote. How often you showed up barely registered. Six decades later the arithmetic has inverted. Denomination tells you little; attendance tells you most of what the old labels once did.
The standard way to see this is to track two gaps side by side. The first is denominational: the Democratic share of the presidential vote among Catholics minus the share among Protestants — the spread that defined the Al Smith and John Kennedy era. The second is devotional: the Democratic share among Americans who rarely attend services minus the share among those who attend often — what political scientists came to call the “god gap.” Across the ANES record one falls as the other climbs, and around 1992 they trade places.
Two religious gaps that swap places
Each line is a gap in the Democratic share of the two-party presidential vote, in percentage points. Denominational: Catholics minus Protestants. Devotional: rare attenders minus frequent attenders. Above zero favors Democrats.
The takeaway: the gap that organized the religious vote at mid-century shrank to nothing; the gap that organizes it now roughly doubled. Devotion replaced denomination as the axis that matters.
The crossover is not a story of religion leaving politics. It is a story of re-sorting. Geoffrey Layman, writing at the close of the 1990s, already saw the old denominational alignment dissolving into a divide between the religiously committed and everyone else, organized around a “progressivist–orthodox” cultural axis rather than the historic Catholic–Protestant one.1 A decade later Robert Putnam and David Campbell put a number on it: by the 2000s, how religious you are had become a sharper political dividing line than which tradition you belonged to. At Eisenhower's election, they note, there was essentially no god gap at all.2
Which church you belonged to no longer moves your vote. How often you walk through its doors moves almost everything.
If devotion has genuinely supplanted denomination, the signature should appear inside each tradition, not just across them. It does. Split Catholics, Protestants, and the religiously unaffiliated each into frequent and infrequent attenders, and within every one of them the rare attenders drifted Democratic while the devout held or moved Republican. The attendance gap that barely existed in the 1970s is now the dominant cleavage in all three.
Within every tradition, the pew divides the vote
Democratic share of the two-party presidential vote, by how often respondents attend services, plotted separately for each tradition. Units: percent voting Democratic. Attendance fielded from 1972.
Catholics
Protestants
None / secular
The takeaway: the same wedge — rare attenders moving Democratic, frequent attenders moving Republican — opens inside Catholicism, inside Protestantism, and among the unaffiliated. Attendance, not affiliation, is now the master variable.
Open markers and dashed end-segments mark the 2024 attendance points, which are upward-biased (see note). Thin cells (the few devout “nones,” sparse early years) are dropped below an unweighted n of 40.
Why did the axis rotate? The conventional reading runs from religion to politics: the cultural conflicts that crystallized after the 1960s — abortion above all, later same-sex marriage — fused religious observance to the Republican coalition and drove the secular toward the Democrats.2 Michele Margolis adds the arrow running the other way. For many Americans a firm partisan identity forms first, in early adulthood, and religious involvement bends to fit it: Democrats drift from the pews, Republicans toward them, so that politics ends up shaping religiosity as much as the reverse.3 Either way, the cleavage that results is devotional, and it deepens.
The most recent elections extend the pattern to its logical end. By 2024 the religiously unaffiliated had become the single largest bloc in the Democratic coalition, and roughly seven in ten of them backed the Democratic nominee; among regular attenders the Republican advantage was just as lopsided.4 What looks from a distance like secularization erasing religion's political force is, up close, the opposite: religion is more politically structuring than ever — only now the structuring variable is the frequency of devotion rather than its label. The denominational vote that organized American elections from Al Smith to the Reagan realignment has been quietly retired, and a devotional vote has taken its place.
Data. ANES Cumulative File, presidential-election years 1948–2024, weighted with the ANES survey weight. The outcome is the Democratic share of the two-party presidential vote among reported voters (non-voters excluded). Denomination uses the ANES four-category religion item (Protestant, Catholic, Jewish, Other/None); the devotional contrast uses the ordered church-attendance item (Rarely / Sometimes / Often), comparing the Rarely and Often categories. The attendance item is first fielded in 1972, so all devotional series begin that year.
Two gaps. The denominational gap is the weighted Democratic vote share among Catholics minus the weighted share among Protestants. The devotional gap is the weighted Democratic share among rare attenders minus the share among frequent attenders. Both are in percentage points; positive favors Democrats. Era figures are simple means over the surveyed years in each window. The crossover (about 1992) is the first surveyed year, excluding the biased 2024 attendance point, at which the devotional gap exceeds the denominational gap and stays above it.
The 2024 attendance caveat. Only about 49% of 2024 ANES respondents have a valid church-attendance value, and the documentation flags regular attenders as over-represented among those who do. The 2024 attendance-based (devotional) points are therefore plotted with open markers and dashed end-segments and are not treated as directly comparable to earlier years; no claim in the text rests on them. The 2024 vote-by-denomination point is unaffected by this problem and is reported normally — including its striking reversal, the first year in the series in which Protestants voted more Democratic than Catholics.
Cell sizes. A year is reported only with at least 200 presidential voters; a denominational or attendance point requires at least 60 cases, and a within-tradition attendance point at least 40. The small group of frequently attending “nones” falls below these thresholds in several early years and is dropped there rather than shown as noise.