Has Partisanship Subsumed Religion? Reassessing Religious Effects on School Prayer in U.S. Politics

Abstract

Religion and partisanship remain deeply intertwined in contemporary American politics, especially in public debates on religious expression in state institutions. This study examined whether religious identity and behavior continue to influence public attitudes independently of party affiliation in a highly polarized environment. Drawing on the latest 2023–2024 Pew Religious Landscape Study, the analysis examined support for teacher-led Christian prayer in public schools—a constitutionally contentious issue—through survey-weighted logistic regression models. The models included key religious predictors—tradition, born-again identity, and church attendance—alongside controls for political ideology and party identification. While Republican partisanship is the single strongest predictor of support, religious identity retains a significant and independent effect. Evangelical Protestants, as well as highly observant individuals across traditions, consistently show greater support for school prayer than their less religious or differently affiliated co-partisans. These residual effects point to the persistence of religious subcultures within each party coalition. By identifying such within-party variation, this study contributes to broader debates on the evolving boundaries of secular governance and the complex interplay between religion and partisan identity.

1. Introduction

American politics has undergone a profound realignment of social cleavages in recent decades. Traditional religious divides that once transcended partisan boundaries have increasingly been subsumed by partisan polarization (J. C. Green 2007). Evangelical Protestants, for example, have become a core Republican constituency, while the religiously unaffiliated (“nones”) and many non-evangelical Christians lean Democratic (Perry 2022). This evolving alignment indicates that partisan identity has become the principal axis structuring political differences that were once predominantly shaped by religious affiliation (Hout and Fischer 2014). The central puzzle motivating this study was whether religion still exerts an independent influence on political attitudes after accounting for the powerful sorting effect of partisanship. More specifically, has partisan polarization absorbed most religious cleavages, or do residual effects of religious identity persist even within partisan camps?

This study hypothesized that partisan polarization has absorbed the bulk of America’s religious–political divisions, aligning most devout religious constituencies with one party and more secular groups with the other. However, it was also theorized that within these largely secularized partisan groupings, religious identities continue to produce meaningful differences in attitudes (G. C. Layman 1997). Even among avowed partisans—for example, among Democrats—those with strong religious commitments may diverge from their co-partisans who are more secular (Djupe 2014). To test this hypothesis, this study examined public support for teacher-led prayer in public schools—a classic culture war issue that has historically divided Americans along religious lines (Dill and Hunter 2010).

Teacher-led devotional exercises were once commonplace in U.S. public schools. Archival surveys catalogued by S. K. Green (2012) indicate that as late as 1958, twenty-nine states either required or explicitly encouraged daily Bible reading or prayer. Litigation by religious minorities and secular groups after World War II culminated in Engel v. Vitale (1962) and Abington v. Schempp (1963), decisions that barred state-sponsored prayer, but left voluntary, student-initiated prayer intact (S. K. Green 2012). Partisan stakes emerged more slowly: both parties initially accepted the rulings, but by 1980 the Republican platform had endorsed a constitutional amendment to restore school prayer, marking the start of a durable GOP–Democratic cleavage (S. K. Green 2015). Subsequent Congresses saw dozens of prayer-amendment bills—all backed predominantly by Republican members—entrenching a partisan cleavage that persists today. This historical arc situates our analysis: what began as a widely shared civic ritual is now a symbolic marker at the intersection of religion and party politics.

By estimating a series of nested logistic regression models, the explanatory power of religious identity was compared to that of partisan affiliation. If partisan polarization has fully absorbed religious cleavages, the inclusion of party identification should substantially attenuate or eliminate the effect of religious tradition and devotion. Conversely, if residual religious effects persist, it is expected that even within partisan groups, an individual’s religious tradition and intensity will continue to shape attitudes toward school prayer.

In pursuing this inquiry, the analysis makes both empirical and theoretical contributions. Empirically, the research leverages the newly released 2023–2024 Pew Religious Landscape Study—a nationally representative survey of over 36,000 U.S. adults—to provide up-to-date evidence on religion and polarization. The large sample and detailed measures of religious affiliation and practice enable granular analysis of even small religious subpopulations. The survey’s timing (conducted in 2023–2024) also permits assessment of these dynamics in the contemporary era of high partisan polarization. Theoretically, this work builds on and speaks to two important literatures. The first is the long tradition of research on religion and political behavior, which has documented how religious affiliations (e.g., evangelical vs. mainline Protestant) and practices influence voting and policy attitudes (G. C. Layman 1997). The second is the literature on partisan sorting and identity alignment, which contends that contemporary American partisanship increasingly consolidates multiple social identities—including religion—into two opposing mega-identities (Klein 2020). By integrating these perspectives, the analysis investigated whether religious identity still matters once partisan identity has effectively structured the religious–secular divide. The answer has important implications for the study of identity politics: if partisan identity truly subsumes religious identity, political conflict may be framed purely in partisan terms. However, if religious identity retains an independent influence, intra-party heterogeneity in political attitudes rooted in religion may persist within party coalitions.

The subsequent analysis unfolds across several key sections. Initially, a theoretical framework emerges through synthesizing existing scholarship on religious identities and partisan polarization to establish testable hypotheses. Subsequently, the article introduces the dataset, variable operationalization, and analytical methods, particularly focusing on survey-weighted logistic regression modeling. The analysis proceeds by comparing nested models—a baseline demographic specification, a religion-only model, a partisanship-only model, and a comprehensive model—to systematically evaluate how accounting for partisanship affects observed religious influences. The concluding discussion examines how these empirical findings illuminate the evolving intersection between religious identity and partisan alignment within contemporary American political contexts.

2. Theoretical Framework

2.1. Religion, Identity, and Realignment

Religion has long constituted a foundational cleavage in American political behavior (Wald 2014). Throughout much of the 20th century, denominational affiliation (e.g., Catholic, Protestant, Jewish) and the intensity of religious observance significantly shaped patterns of electoral behavior and policy preferences (Manza and Wright 2003). Beginning in the 1970s and accelerating through the 1990s, the major parties experienced a “religious realignment,” sorting increasingly along religious lines (Brooks and Manza 2004). The Republican Party became closely associated with evangelical Protestants and conservative Catholics—groups often mobilized by moral and cultural issues such as abortion and school prayer—while the Democratic coalition expanded to include religious minorities, secular individuals, and mainline Protestants with more liberal social values (Dionne 2016). This realignment, marked by the rise of the “religious right” and a parallel secularization of the political left, produced what Putnam and Campbell (2010) describe as a “God gap,” a widening partisan divide between the devout and the less religious. Over subsequent decades, frequent churchgoers and self-identified born-again evangelicals increasingly gravitated toward the GOP, while individuals with no religious affiliation (“nones”) or more pluralistic religious outlooks moved toward the Democratic Party (Margolis 2018). The early 2000s marked a period when religious affiliation and observance intensity emerged as dominant predictive factors in determining partisan loyalty and electoral behavior (J. C. Green 2007).

Crucially, this religious realignment unfolded alongside—and contributed to—the broader process of partisan polarization (J. J. Castle and Stepp 2021). As the major parties became increasingly sorted along religious and other social identity lines (e.g., race, ideology), the sociological overlap between Democratic and Republican constituencies diminished significantly (Brooks and Manza 2004). As Mason (2018) argues, partisan identities in the U.S. have evolved into “mega-identities”—bundles of overlapping social characteristics, among which religious affiliation is a critical component. Contemporary Republican voters are disproportionately likely to also be an evangelical Christian (or at least a frequent churchgoer) and to hold orthodox religious views, whereas a Democrat is more likely to be secular or less religiously observant (Hertzke 2019). This social sorting heightens partisan conflict because differences that were once cross-cutting are now reinforcing (Hetherington and Weiler 2009). In earlier eras, churchgoing Democrats and secular Republicans were relatively common, allowing for more complex intersections between religious and partisan identities (Margolis 2018). This consolidation of religious and partisan identities represents a fundamental shift from the cross-cutting cleavage structure that historically characterized American politics (G. Layman 2001). Today, however, such overlaps are far less common. Religious progressives and secular conservatives now represent only small and shrinking segments of the electorate (Abramowitz and Saunders 2008).

As a consequence of this sorting process, partisan identity may now subsume cleavages that were once primarily religious in nature. The theory of conflict extension (G. C. Layman and Carsey 2002) posits that emerging issue cleavages—particularly those concerning moral–religious values—were incorporated into the partisan divide, rather than remaining orthogonal to it. As a result, voters adjusted their partisan affiliations to better reflect their cultural and religious commitments, thereby contributing to increased intra-party homogeneity. Over time, the Republican Party came to predominantly represent a traditionalist religious agenda, emphasizing public expressions of Christianity, opposition to abortion, and related moral stances, while Democrats became the party of church–state separation and religious pluralism (Wilcox 2018). Consequently, party identification now reveals far more about an individual’s religious orientation than in previous eras and vice versa. Public opinion research confirms that many Americans consciously link religion with partisanship—for example, associating evangelicals with the GOP and secular individuals with the Democratic Party (Campbell et al. 2018). Taken to its logical extreme, this dynamic suggests that the influence of religion on political attitudes may increasingly be mediated by partisan identity such that political disagreement is framed not as Protestant versus Catholic, but as Republican versus Democrat (Patrikios 2008).

2.2. Residual Religious Effects Within Parties

Despite this substantial partisan–religious alignment, there remain compelling reasons to expect that religion continues to shape political attitudes in more nuanced, intra-party ways. First, not all religious differences correspond directly to the two-party system. For example, within American Protestantism, evangelical and mainline Protestants now largely differ by party: evangelicals mostly Republican, mainliners often Democratic or moderate (Howe 1991). But consider historically Black Protestants: this group is theologically evangelical in many respects, yet predominantly Democratic in party identification due to racial and historical factors. African American churchgoers frequently espouse traditional religious values akin to those of White evangelicals, such as strong support for public prayer, but they remain a key Democratic constituency (Harris 1999). This creates an instructive example: within the Democratic Party, Black Protestants might be more supportive of public prayer or other religious expressions than their secular/liberal Democratic peers (Leon McDaniel et al. 2011). Similarly, within the GOP, there are religious minorities and less observant Christians who identify as Republican for economic or cultural reasons, but do not share the intense religiosity of the party’s evangelical base. These intra-party differences suggest that when partisanship is held constant, variation in religiosity and religious tradition could still yield meaningful attitudinal differences (Jelen 1998). In short, partisan sorting has not made the parties religiously monolithic: there is still diversity within each party’s ranks, albeit less than in the past. Evidence of this diversity is expected to manifest in policy attitudes that are closely tied to religion.

Second, research on political socialization and identity suggests a dynamic, bidirectional relationship: while many individuals adapt their partisan identities to reflect their religious commitments, others recalibrate their religious engagement to better align with their political preferences (Margolis 2018). For instance, an individual with evangelical convictions who identifies as a Democrat may experience cognitive dissonance in today’s highly polarized partisan environment: some respond by reducing their religious involvement, whereas others retain both identities despite the tension. Those who maintain a strong religious identity within a largely secular political camp might display distinctive attitudes (Egan 2020). Whitehead and Perry (2020) documented that the rise of the Christian right in the GOP prompted some liberals to become disaffiliated from religion, but those religious liberals who stayed could be quite devout. Thus, a devout Democrat might still differ from a secular Democrat on issues like prayer in schools, even if they share a party. Conversely, secular Republicans—though increasingly uncommon—may exhibit less support for religiously conservative policies than their evangelical counterparts within the party.

From the perspective of identity theory, individuals hold multiple social identities concurrently. While partisan identity is often highly salient, it does not necessarily displace other salient identities. Instead, individuals tend to activate or prioritize specific identities depending on contextual cues (Huddy 2001). In contexts characterized by heightened partisan salience—such as electoral campaigns—partisan identity may become dominant (Ruckelshaus 2022). However, when individuals evaluate policy issues with explicit or implicit religious content, their religious identity may exert influence—even if only subtly (Tajfel and Turner 2004). The notion of “complex religion” (Wilde 2018) also suggests that religious identity cannot be reduced to affiliation alone: factors like personal piety (prayer frequency, scripture belief) cut across denominational lines. These dimensions of religiosity may operate meaningfully within partisan groups. For instance, among Democrats, frequent churchgoers may be more supportive of faith-accommodating policies than their non-attending co-partisans, despite their shared partisan affiliation (Huddy et al. 2015).

2.3. Expectations

Synthesizing these theoretical insights yielded the following expectations. (1) Bivariate differences in support for school prayer between major religious traditions (e.g., evangelical Protestants versus others) will be substantial, but will largely reflect underlying partisan alignment. Evangelical Protestants were expected to express high support for teacher-led school prayer, consistent with their predominantly Republican affiliation. Religious “nones” or less observant mainline Protestants will show low support and lean Democratic (Bolyen, #10). In multivariate models, the inclusion of partisan identity was expected to attenuate the estimated effects of religious tradition and individual religiosity. Much of the effect of being evangelical, for instance, would be accounted for by the fact that evangelicals are Republican. Statistically, religion variables that are significant predictors in a model with only demographics (M-0) are expected to decrease in magnitude or statistical significance once party is controlled (in M-3), illustrating the extent to which polarization absorbs the religious cleavage. (3) However, the influence of religion was not expected to be entirely mediated by partisanship. Certain religious indicators—such as born-again identification or frequency of church attendance—were anticipated to remain statistically significant in the full model, indicating residual religious effects net of partisanship. Specifically, evangelical Protestant affiliation was expected to retain an independent positive effect on support for school prayer even after controlling for party such that, for example, an evangelical Democrat is more supportive of school prayer than a similarly situated non-evangelical Democrat (and likewise within the GOP). Measures of personal religiosity (attendance and born-again self-identification) were also expected to continue to show effects, reflecting within-party intensity differences. In sum, while partisan polarization may have compressed inter-party differences to dominate overall variance on religious issues, it had not eliminated intra-party heterogeneity altogether. The empirical question is the magnitude of these residual effects and whether they are substantively meaningful.

By exploring these questions, this study engages with a growing body of work arguing that American politics is now essentially a tribal conflict between a mostly religious right and a mostly secular left. This study’s contribution lies in examining the religious and secular minorities within each partisan “tribe”—the secular conservatives and religious liberals—and measuring their impact. The findings will reveal whether the partisan divide has completely subsumed religious cleavages or whether American political behavior still needs to account for religion as a distinct force.

3. Data and Methods

Data—Pew 2023–2024 Religious Landscape Study: This study draws on data from the 2023–2024 Pew Research Center Religious Landscape Study (RLS), a nationally representative survey of U.S. adults (N = 36,908). The RLS constitutes a comprehensive nationwide study examining religious patterns across the American population, collecting detailed information on respondents’ religious affiliations, beliefs, and practices, alongside a range of sociopolitical attitudes. Crucially, the survey’s large sample facilitates robust analysis of relatively small religious subgroups (e.g., Mormons, Orthodox Christians), offering greater statistical precision than is typical in nationally representative polling. Data were collected from July 2023 to March 2024 using a multimodal approach comprising online, paper-based, and telephone surveys. The data were weighted to account for the survey’s complex sampling design and differential nonresponse, ensuring representativeness of the U.S. adult population (Pew 2025a). The analytic sample included respondents from the main survey—excluding a small experimental subsample—with complete data on all variables of interest. Following listwise deletion for item nonresponse (described below), the final analytical sample consisted of 22,706 respondents, exceptionally large by national survey standards and affording considerable statistical power. All analyses applied the provided sampling weights and replicated weights to ensure unbiased point estimates and correct standard errors.

Dependent Variable: The dependent variable in this analysis was support for teacher-led prayer in public schools. The survey included two related items measuring attitudes toward school prayer: (1) whether they favor or oppose allowing public school teachers to lead students in prayers that refer to Jesus (i.e., explicitly Christian prayer), and (Bolyen, #10) favor or oppose teachers leading prayers that refer to God in general, without mentioning a specific religion. Response options for each were on a four-point scale from “strongly favor” to “strongly oppose.” For this analysis, a binary indicator PRAYER_SUPPORT was derived, equal to 1 if a respondent favored or strongly favored Christian prayer led by teachers, and 0 if the respondent opposed or strongly opposed it. Essentially, PRAYER_SUPPORT measured approval of the most explicit form of school prayer (one referencing Jesus). This stricter operationalization was employed for two primary reasons: it presents a clearer test of attitudes toward the church–state boundary (state-employed teachers leading sectarian prayer), and it serves as a conservative test of religious influence, given that support for explicitly Christian school prayer is more likely to reflect underlying religious conviction. In the analytical sample, 52% of respondents expressed support (either “favor” or “strongly favor”) for this practice, a level consistent with the Pew Research Center’s published findings and broadly aligned with prior surveys reporting majority support for some form of school prayer. The binary coding simplified the interpretation of a “pro-prayer” versus “anti-prayer” stance while maintaining substantive meaning. Robustness checks using alternative specifications, including ordinal logistic regression with the full 4-point scale and broader coding that included support for generic, but not sectarian prayer, yielded substantively similar results.

Key Independent Variables—Religion: To capture the multidimensional nature of religious identity and practice, the analysis incorporated three widely used measures: religious tradition, born-again identification, and attendance frequency.

Religious tradition was measured using Pew’s RELTRAD classification, a widely adopted typology that aggregates denominational data into broader religious traditions (Steensland et al. 2000). A series of dummy variables was constructed for the following major Christian traditions: evangelical Protestant, mainline Protestant, historically Black Protestant, Catholic, Mormon (i.e., Church of Jesus Christ of Latter-Day Saints), Orthodox Christian, and an “other Christian” category that combined smaller groups (e.g., Jehovah’s Witnesses and other minor denominations). Because the survey item explicitly referred to Jesus, non-Christian respondents were far more likely to choose “Don’t know” (37 percent versus < 4 percent among Christians). To avoid conflating item nonresponse with substantive attitudes, the main analysis was limited to Christian traditions, with a full-sample robustness check provided in Appendix A. In all regression models, mainline Protestants—including historically Black Protestants—were designated the reference category. This choice of reference provided a useful baseline: mainline Protestants have traditionally been moderate in both religiosity and politics (Putnam and Campbell 2010), and including Black Protestants in the reference group acknowledged that despite different racial composition, historically Black churches often occupy a theologically mainline position (in terms of not emphasizing “born-again” evangelical theology). In practice, combining mainline and Black Protestants yields a reference group of “non-evangelical Protestants,” against which evangelicals, Catholics, and other groups are compared. Models with separate categories for mainline and Black Protestants were also estimated. Detailed estimates with the two Protestant traditions separated appear in Appendix A. The effects were in the same direction, with Black Protestants somewhat more pro-prayer than White mainline Protestants. For reasons of parsimony, these groups are merged in the models presented here.

Beyond affiliation, the analysis incorporated BORN-AGAIN identification (variable BORNFINAL), which measured whether a Christian respondent describes themselves as a “born-again or evangelical Christian.” This item was originally coded 1 for yes and 2 for no, but recoded as a binary indicator (1 = born-again, 0 = non-born-again) for analysis. Non-Christian respondents were not asked this item and were coded as missing on BORNFINAL. Including this variable enabled the capture of an element of religious identity and theology that cut across denominations: for instance, a Catholic or mainline Protestant who considered themselves “born-again” (a minority in those groups) versus one who did not. This was highly correlated with evangelical Protestant affiliation (virtually all White evangelicals, by definition, respond affirmatively), but provided additional information for other groups. BORNFINAL was included to assess whether individual-level evangelical orientation exerts an independent effect on policy attitudes, net of denominational affiliation.

Religious attendance frequency (CHATTEND) was included as a key behavioral indicator of religiosity (Djupe and Gilbert 2008). The variable was ordinal, ranging from 1 (“more than once a week”) to 6 (“never”), with intermediate categories reflecting decreasing frequency of attendance. In regression models, CHATTEND is treated as an ordinal variable with a linear functional form, effectively assuming a constant effect of one-unit increases on the log-odds of supporting school prayer. For ease of interpretation, the variable is reverse-coded in descriptive analyses so that higher values indicate greater religiosity; however, in the regressions, the original coding is retained (so a negative coefficient on attendance means more frequent attendance—lower numeric code—and is associated with higher prayer support). Church attendance is a classic measure of religious commitment and social exposure to religion, and numerous studies have found it predictive of political attitudes on culture-war issues (Galen 2012). Including attendance enabled an assessment of whether intensity of religious practice matters net of mere affiliation. It also helped capture differences within a tradition, e.g., a never-attending evangelical vs. a weekly-attending evangelical might have different views on public prayer.

In sum, Model M-1 (Religion model) included religious tradition dummies (evangelical, Catholic, Mormon, Orthodox, other Christian; mainline/Black Protestant as reference), born-again (yes/no), and attendance (ordinal), in addition to a baseline of control demographics (described below). These covered both identity-based aspects of religion (affiliation and born-again self-concept) and behavioral aspects (attendance).

Key Independent Variable—Partisanship: The other crucial independent variable was party identification. The Pew survey’s PARTY item asked respondents to identify as Republican, Democrat, Independent, or “Other.” Dummy variables were constructed for Republican, Independent, and other affiliations, using Democrat as the reference category in all models. Respondents who initially chose “Independent” or “Other,” but expressed a partisan leaning in follow-up questions are retained in their original category. This approach reflected the decision to measure self-identified group membership rather than inferred partisan leanings. Robustness checks using a recoded three-category scheme that reclassified leaners as partisans yielded substantively similar results. Respondents who refused to report a party affiliation (approximately 3.5% of the sample) were excluded from the analytic sample to maintain clarity and interpretability. In Model M-2 (Party), these partisan dummy variables were included alongside demographic controls. This model estimated the extent to which partisanship alone predicts support for school prayer.

The full Model M-3 (Religion + Party) included all the above religious variables and party dummies together (plus controls). Comparing Models M-1, M-2, and M-3 allowed for observation of how the estimated effects of religious identity are attenuated or sustained once partisanship is included and vice versa. Consistent with the theoretical framework, it was anticipated that partisanship would explain a substantial share of the variance in support for school prayer, but that meaningful residual effects of religious identity would remain.

Control Variables: All models incorporated a baseline set of demographic controls (Model M-0) designed to account for individual-level characteristics that may independently influence attitudes toward teacher-led school prayer. Gender was coded 1 for female and 0 for male. Racial and ethnic identity was operationalized using dummy variables for Black, Asian, multiracial, and other racial groups, with non-Hispanic White as the reference category. Hispanic respondents were incorporated into these racial categories based on self-identification. The “other” category included Native American and respondents not otherwise classified. Educational attainment was measured by four categories: high school or less (reference), some college (including associate’s degree), bachelor’s degree, and postgraduate degree. Given that higher educational attainment is often associated with more liberal views on church–state issues, controlling for education helped disentangle religious effects from educational confounds. Finally, political ideology (Bolyen, #10) was included, measured on a 5-point scale from very conservative (1) to very liberal (5). This served two purposes: (a) as a control for general left–right orientation, since support for school prayer might reflect broader ideological tendencies not entirely captured by party or religion, and (b) as a rough proxy for any remaining value differences. Ideology was treated as a continuous variable in the models. Although party identification and ideology are strongly correlated, both were deliberately included in the full specification to isolate the distinct contribution of religious identity beyond general ideological orientation. In practice, including ideology did not change the substantive patterns of interest, but it absorbed some variance in the baseline model (M-0), thereby providing a more conservative test of the effects of religion and partisanship.

Estimation—Survey Logit with Replicate Weights: Given the dependent variable was binary, logistic regression models were estimated using maximum-likelihood estimation (MLE). To account for the complex sampling design of the Pew survey, all models incorporated survey weights. Specifically, the analysis employed survey settings equivalent to Stata 19 specifications: the person-level weight provided by Pew (WEIGHT) as the probability weight, and 500 replicate weights (REPWT_001 to REPWT_500) with Fay’s balanced repeated replication (BRR) variance estimation method (Fay coefficient ρ = 0.5). This approach adjusted for the survey’s stratified and weighted design, yielding valid standard errors and statistical inferences. Coefficient estimates were derived from the weighted likelihood, and standard errors were computed from replicate weights. For robustness, estimates were cross-verified using Taylor series linearization, which produced substantively identical results. All p-values reported (e.g., p < 0.05) were based on two-tailed hypothesis tests using replicate-weight-adjusted standard errors to ensure conservative inference.

It is worth noting that with the large sample, even minimal differences may achieve statistical significance. Therefore, effect sizes and substantive significance are emphasized over mere statistical significance. To enhance interpretability, average marginal effects (Collaborators, #3) are presented, which translate the logistic coefficients into percentage-point changes in predicted probability of supporting school prayer. For categorical (dummy) variables, the AME represents the difference in predicted probability between the focal category (e.g., evangelical Protestants) and the reference group (e.g., mainline Protestants), holding all other covariates at their observed values. For continuous variables, the AME indicates the change in predicted probability associated with a one-unit increase in the independent variable. In the case of political ideology, this corresponds to the difference between adjacent points on the five-point ideological scale.

All hypothesis tests were conducted at conventional levels. In tables, asterisks are used to denote significance (* p < 0.05, ** p < 0.01, *** p < 0.001). While odds ratios are occasionally reported in-text for interpretive clarity, all regression tables present untransformed logit coefficients to preserve additivity and comparability across nested model specifications.

By examining the nested models, the change in model fit can also be observed. Both M-1 (adding religion) and M-2 (adding party) yielded substantial increases in pseudo-R² relative to the baseline model, with the full model (M-3) exhibiting the highest explanatory power. These improvements are interpreted as evidence of each factor’s contribution.

In summary, the modeling strategy evaluated whether explanatory power shifted from religious variables to partisan identity across models. If partisan polarization has fully absorbed religious cleavages, religion was expected to add little explanatory value once party was included. Conversely, if residual religious effects persist, coefficients for religious tradition and practice were anticipated to remain statistically significant in M-3 and model fit would improve relative to M-2.

Before presenting the regression models, key descriptive patterns in the data are briefly examined to establish baseline associations. Overall, 52% of respondents expressed support for teacher-led Christian prayer. Bivariate tabulations show stark divides. Support varies sharply by partisanship: 85.1% of self-identified Republicans favor it compared to 50.3% of Democrats—a 34.8-point gap. By contrast, among religious traditions, 84.9% of evangelical Protestants express support compared to 59.2% of mainline Protestants—a 30.4-point difference. Attendance also predicts attitudes: 81.2% of weekly attenders support the practice versus 42.3% of those who never attend—a 38.9-point difference (Pew 2025a). These descriptive patterns underscore strong bivariate associations, but multivariate models are needed to disentangle the independent contributions of religion and partisanship, especially given their strong correlation. For instance, given that most evangelicals identify as Republicans, to what extent is the evangelical–non-evangelical divide merely a reflection of partisan alignment? The multivariate analysis addresses these questions directly.

4. Results

Model Comparisons (

Table 1. Nested logistic models of support for teacher-led prayer in public schools.

	M-0 (Collaborators, #3)	M-1 (+ Religion)	M-2 (+ Party)	M-3 (Full)
A. Demographics
Female (1 = F)	−0.122 (0.024) ***	−0.105 (0.023) ***	−0.113 (0.023) ***	−0.108 (0.022) ***
Black vs. White	0.231 (0.033) ***	0.146 (0.032) ***	0.229 (0.033) ***	0.231 (0.031) ***
Asian vs. White	−0.402 (0.145) **	−0.311 (0.140) *	−0.397 (0.143) **	−0.296 (0.139) *
Latino vs. White	0.087 (0.052) †	0.046 (0.050)	0.081 (0.051)	0.044 (0.049)
Multiracial/other	0.064 (0.061)	0.012 (0.060)	0.061 (0.061)	0.014 (0.059)
Some college	−0.214 (0.028) ***	−0.206 (0.027) ***	−0.205 (0.028) ***	−0.198 (0.027) ***
Bachelor’s deg.	−0.553 (0.041) ***	−0.512 (0.040) ***	−0.494 (0.041) ***	−0.489 (0.039) ***
Post-grad deg.	−0.802 (0.059) ***	−0.735 (0.057) ***	−0.718 (0.058) ***	−0.712 (0.056) ***
B. General ideology/values
Ideology (1 cons.→5 lib.)	−0.704 (0.022) ***	−0.618 (0.022) ***	−0.442 (0.023) ***	−0.395 (0.022) ***
C. Religious tradition
Evangelical Prot.	—	1.10 (0.05) ***	—	0.67 (0.07) ***
Catholic	—	0.47 (0.06) ***	—	0.08 (0.07)
LDS	—	0.96 (0.17) ***	—	0.59 (0.15) ***
Orthodox Chr.	—	0.30 (0.19)	—	0.14 (0.18)
Other Christian	—	0.22 (0.09) *	—	0.11 (0.09)
D. Religious devotion/practice
Born-again (ID = 1)	—	0.59 (0.04) ***	—	0.41 (0.05) ***
Attendance (↑ = less)	—	−0.21 (0.03) ***	—	−0.17 (0.03) ***
E. Party identification
Republican vs. Dem.	—	—	1.36 (0.07) ***	1.15 (0.08) ***
Independent vs. Dem.	—	—	0.32 (0.06) ***	0.15 (0.08) †
Other party vs. Dem.	—	—	0.45 (0.23) *	0.34 (0.21)
Model-fit indices
Pseudo R2	0.039	0.094	0.086	0.172
AIC	26 480	24 830	24 970	23 015
LR χ2 (df)	—	1 795 (14) ***	1 834 (6) ***	2 365 (20) ***

* Weighted survey logit, N = 22,706; reference categories: Male/White/≤HS/Mainline-Black Prot./Democrat; SE in parentheses; *** p < 0.001, ** p < 0.01, * p < 0.05, ^† p ≈ 0.05; McFadden pseudo-R² reported. For AIC and LR χ², design-based F-tests (500 replicate weights, Fay ρ = 0.5) were used. Attendance is coded with higher values indicating less frequent attendance; thus, a negative coefficient means more frequent attendance increases support for school prayer.

): The analysis begins with nested logistic regressions predicting support for school prayer (PRAYER_SUPPORT). Table 1 presents the coefficients from four models: M-0 with only demographics and ideology, M-1 adding religious factors, M-2 adding party ID (instead of religion), and M-3 including all predictors. This allows a stepwise evaluation of the hypothesis.

Baseline controls (Model M-0). Several demographic and ideological factors’ structure support for teacher-led prayer even before religion or party is introduced. Women are modestly less supportive than men (Coef. = −0.122, p < 0.001), echoing evidence that women, despite higher religiosity, often favor stricter church–state separation. Black respondents are more supportive than Whites (Coef. = +0.231, p < 0.001), consistent with the historic centrality of the Black church in public life. Latinos and multiracial respondents also trend pro-prayer, whereas Asian Americans are less supportive (Coef. = −0.402, p < 0.01). Education exerts a strong, graded negative effect: college graduates are far less supportive than those with a high-school education or less (bachelor’s = −0.553; master’s+ = −0.802; both p < 0.001), mirroring the link between higher education and secular orientations. Finally, ideology is potent: each one-unit shift toward liberalism lowers support (Coef. = −0.704, p < 0.001): a “very liberal” American is dramatically less inclined to endorse school prayer than a “very conservative” one. These patterns show that even absent religion or party, cultural outlooks and sociodemographics strongly predict attitudes on prayer in schools.

Moving to Model M-1, which adds religious tradition, born-again identification, and church attendance to the baseline specification, a meaningful increase in explanatory power is observed (pseudo-R² rises from 0.039 to 0.094). Among religious variables, evangelical Protestants exhibit significantly higher support for teacher-led school prayer compared to mainline Protestants, even after controlling for demographics and ideology. Latter-Day Saints also show strong positive effects, whereas the coefficient for Catholics does not reach statistical significance. Born-again identification is independently associated with higher support, and more frequent church attendance likewise predicts greater favorability toward school prayer. Taken together, these findings highlight that both religious affiliation and personal religiosity—measured here by church attendance—contribute to shaping attitudes toward public expressions of faith beyond what is explained by demographics or ideology alone.

Adding religion-related variables in M-1 alters the interpretation of some baseline controls. For instance, the gender gap slightly attenuates: the female coefficient moves from −0.122 in M-0 to −0.105 in M-1, suggesting that women’s lower support is partially explained by their greater religiosity on average. Similarly, the coefficient for Black respondents drops from +0.231 to +0.146 (both p < 0.001), indicating that a significant portion of their higher support for school prayer stems from higher levels of religious observance (e.g., attendance, born-again identity). Education and ideology effects remain strong and statistically significant, though their magnitudes slightly shrink, reflecting partial overlap with religiosity: college-educated and liberal individuals tend to be less religious, which religion controls help disentangle. These patterns reinforce that while demographic and ideological divides persist, religious identity and behavior are central to understanding public opinion on prayer in schools. Absent party controls, the U.S. public appears divided along religious lines, a finding consistent with scholarship on religious cleavages in cultural policy debates (J. Castle 2019).

Model M-2 introduces party identification while holding ideology constant, but excluding all religious affiliation or practice variables. The inclusion of partisan alignment markedly improves model fit, with the McFadden pseudo-R² rising from 0.039 in the baseline Model M-0 to 0.086. Partisanship emerges as a strong predictor: respondents identifying as Republicans are significantly more likely to support teacher-led prayer than Democrats (b = +1.36, SE = 0.07, p < 0.001), and Independents also show greater support than Democrats (b = +0.32, p < 0.001). Those affiliated with “other” parties display a weaker, but still statistically significant positive effect (b = +0.45, p ≈ 0.05). These results underscore the centrality of partisan identity in shaping public attitudes on church–state issues. Notably, the addition of party attenuates some covariate effects: the ideology coefficient shrinks from −0.70 in Model M-0 to −0.44 in M-2 (both p < 0.001), indicating that part of the ideological divide observed earlier reflects underlying party alignment. Education and gender effects remain directionally stable with only slight reductions in magnitude, while the coefficient for Black respondents continues to be positive and statistically significant (b = +0.23, p < 0.001). This enduring effect suggests the presence of racial heterogeneity within partisan groups, an issue explored more closely in the next model.

Model M-3 incorporates both religious and partisan predictors, offering a more complete account of support for teacher-led school prayer. Compared to earlier models, M-3 shows that partisan alignment explains a substantial share of the variation, but religious identity and practice retain significant effects. Evangelical Protestants continue to stand out: even after controlling for partisanship and other covariates, their support remains significantly higher than that of mainline Protestants (b = +0.67, p < 0.05). By contrast, the positive effect for Catholics observed in M-1 becomes statistically insignificant, suggesting that their prior distinctiveness was largely a function of party alignment. Mormons remain highly supportive of teacher-led prayer (b = +0.59, p < 0.001), indicating a uniquely strong religious stance not entirely explained by political affiliation.

In terms of religious behavior and identity, both frequent church attendance (b = −0.17, p < 0.05; where lower values indicate higher frequency) and born-again self-identification (b = +0.41, p < 0.05) remain statistically significant predictors, pointing to the enduring relevance of personal religiosity beyond institutional affiliation. Republican partisanship continues to exert the strongest influence overall (b = +1.15, p < 0.001), though the effect size slightly declines from M-2. In sum, partisan sorting explains much of the evangelical–mainline gap, but a distinct “evangelical premium,” along with effects of attendance and born-again identity, endures within parties. Religious practice and conviction therefore continue to differentiate citizens even after the contemporary party system has partially absorbed the broader denominational cleavage.

To put these effects in perspective, attention turns to the substantive interpretations in

Table 2. Average marginal effects (AMEs) on the probability of favoring teacher-led Christian prayer.

Predictor (Reference → Level)	Δ P (pct-pts)	95% CI
A. Demographics
Female (vs. Male)	−1.6 *	−3.2, 0.0
Black (vs. White)	+2.3 **	+0.8, 3.8
Asian (vs. White)	−6.8 **	−11.5, −2.1
Latino (vs. White)	+1.1	−1.6, 3.8
Multiracial/Other (vs. White)	+0.4	−2.8, 3.5
Some college (vs. ≤ HS)	−4.2 ***	−6.4, −2.0
Bachelor’s degree + (vs. ≤ HS)	−10.9 ***	−13.2, −8.6
Post-graduate degree (vs. ≤ HS)	−15.0 ***	−18.7, −11.3
B. Ideology
Ideology (+1-unit liberal)	−8.9 ***	−9.8, −7.9
C. Religious tradition
Evangelical Protestant (vs. Mainline/Black Prot.)	+15.6 ***	+12.4, 18.8
Catholic (vs. Mainline/Black Prot.)	+2.0	−0.7, 4.7
Latter-Day Saint (vs. Mainline/Black Prot.)	+8.9 ***	+3.9, 13.9
Orthodox Christian (vs. Mainline/Black Prot.)	+3.2	−2.3, 8.7
Other Christian (vs. Mainline/Black Prot.)	+3.0	−3.4, 9.3
D. Religious devotion/practice
Born-again ID (yes vs. no)	+10.3 ***	+7.5, 13.1
Weekly attendance (vs. never) †	+16.4 ***	+12.2, 20.6
E. Party identification
Republican (vs. Democrat)	+28.3 ***	+25.6, 31.0
Independent (vs. Democrat)	+4.1 **	+1.3, 7.0
Other party (vs. Democrat)	+7.8 *	+0.2, 15.4

* Computed from Model M-3 (full specification); N = 22,706; baseline-predicted PR (support) = 0.58 for the reference profile (male, White, ≤HS, mainline/Black Protestant, Democrat, ideology = 3 “moderate,” never attend). Standard errors are survey design-adjusted (500 bootstrap replicate weights, Fay ρ = 0.5). Effects are expressed in percentage-points; † comparison equals a four-step change in attendance (never → weekly). Δ P = change in predicted probability; values are expressed in percentage points (pct-pts). Significance: * p < 0.05; ** p < 0.01; *** p < 0.001.

, which reports average marginal effects for key variables from the full model (M-3). AMEs translate the logit coefficients into differences in probability (with all other variables averaged out). This helps answer: How much does being evangelical (versus mainline) increase one’s likelihood of supporting school prayer, holding other factors constant? How much more likely is a Republican than a Democrat to support it, controlling for religiosity? And so on.

Table 2 shows that on average, Republican identification increases the probability of favoring teacher-led prayer by approximately 28.3 percentage points compared to Democratic identification (AME = +28.3 ***). This is a remarkably large effect for a single variable in survey research, underscoring how deeply polarized the school prayer issue has become along partisan lines. By contrast, identifying as an Independent raises support by only 4.1 points relative to Democrats (p = 0.004), indicating a modest and much weaker association.

Religious affiliation also continues to matter in meaningful ways. Evangelical Protestants are 15.6 percentage points more likely to support school prayer than mainline and Black Protestants (AME = +15.6 ***), even after controlling for party, ideology, and demographics. This “evangelical premium” is substantial and persists across partisan contexts, and confirms that religious subculture and theological distinctiveness exert influence independently of political alignment.

Self-identifying as born-again further increases support by 10.3 percentage points (AME = +10.3 ***), suggesting that personal religious identity exerts a distinct effect beyond denominational label. Similarly, weekly religious attendance (as opposed to never attending) yields an AME of +16.4 points (computed as Δ between “>1×/week” and “never”), again reinforcing the role of active religiosity in shaping political attitudes. These residual effects suggest that religious conviction and engagement operate within party lines, differentiating attitudes among co-partisans.

Other covariates confirm known patterns. Black respondents, net of religion and party, are about 2.3 points more supportive of school prayer, though the effect size is modest. Gender shows a marginal difference, with women 1.6 points more likely than men to support prayer (p = 0.046). Educational attainment displays a significant negative relationship: those with a bachelor’s degree or higher are 10.9 percentage points less supportive than those with a high school education or less (AME = −10.9 ***).

Taken together, these results support both core hypotheses. Partisan realignment has absorbed a substantial share of the raw religious divide, as evidenced by the reduction in evangelical and born-again coefficients when party is controlled. Yet substantial residual religious effects persist within parties, particularly among evangelical Protestants, born-again Christians, and regular worship attenders. These effects are not reducible to partisanship alone. Thus, both religion and party remain independent and powerful forces shaping public opinion on this high-salience church–state issue. Taken together, these results confirm both core hypotheses: partisan alignment explains a large share of the raw religious divide, but does not eliminate it. Residual religious effects—especially evangelical affiliation, born-again identity, and worship attendance—remain influential within parties. Likewise, partisanship and ideology continue to shape prayer attitudes even among religiously similar individuals. School prayer thus remains a site of intersection between theological conviction and political identity.

To directly examine the presence of residual religious effects within partisan groups, subgroup logit models predicting support for school prayer among Republicans and Democrats separately were estimated, as shown in

Table 3. Evangelical effect within party subsamples.

Subsample	Evangelical Protestant (vs. Mainline/Black Prot.)	OR	95% CI	p
A. Republicans	Evangelical (vs. non-evangelical)	1.06	0.95–1.19	0.28
B. Democrats	Evangelical (vs. non-evangelical)	1.29 *	1.00–1.67	0.048

* Survey-weighted logit estimates with full covariate set (Model M-3). Reference categories inside each subsample: mainline/Black Protestant, non-born-again, never attend, male, White, ≤HS, ideology = “moderate.” OR—odds ratios. Standard errors design-adjusted with 500 bootstrap replicate weights (Fay ρ = 0.5). Significance: * p < 0.05.

. Odds ratios (Collaborators, #3) are reported for evangelical Protestants versus mainline/other Protestants, controlling for full covariates and using replicate weights. Among Republicans, the evangelical effect is statistically indistinguishable from zero (OR = 1.06, p = 0.28), indicating that partisan alignment has largely absorbed the religious cleavage. Evangelical and non-evangelical Republicans exhibit similarly high support for school prayer once other traits are accounted for, suggesting strong partisan consolidation around this issue.

Among Democratic respondents, those affiliated with evangelical Protestant denominations—as classified by the RELTRAD scheme—remain markedly more supportive of teacher-led prayer than mainline or religiously unaffiliated Democrats (OR = 1.29, p = 0.048), even after controls for ideology, education, race, and worship attendance. This coefficient refers to denominational affiliation, not to self-identified “born-again/evangelical” identity, which is modeled separately. RELTRAD places the major historically Black Protestant bodies in a distinct tradition, so they are not counted in the evangelical-affiliation category reported here. By contrast, most Pentecostal churches—including Latino Pentecostal congregations—are coded as part of the evangelical Protestant family, which explains their inclusion in this estimate (see Pew RLS codebook). The result therefore captures the attitudinal effect of belonging to an evangelical denomination within the Democratic coalition independently of racial or ideological factors.

These findings reinforce the dual-structure argument: while partisan realignment has absorbed a large share of the religious divide—especially on the Republican side—religious identity remains salient within parties, particularly among Democrats. The near-neutral effect among Republicans suggests consolidation, while the persistence among Democrats reveals meaningful internal diversity.

5. Discussion

These findings illustrate a contemporary American political landscape in which partisan identity and religious identity are deeply intertwined, but not interchangeable. Partisan polarization has to a great extent absorbed traditional religious cleavages: upon inclusion of party ID in models of support for school prayer, the explanatory power of religious affiliation dropped markedly. This aligns with a large body of research observing that since the late 20th century, religious conservatism has become a pillar of Republican identity, whereas secularism or moderate religiosity characterizes much of the Democratic identity (J. Smith 2023). The analysis reveals a striking partisan divide on this church–state issue: identifying as Republican is associated with a 28.3 percentage-point higher probability of favoring teacher-led prayer compared to identifying as Democrat. This effect magnitude rivals partisan gaps observed on other high-salience issues such as abortion or same-sex marriage (Bankert 2024). In this sense, party has become a shorthand for a cluster of values and identities that include one’s stance on religion in public life. Voters no longer need to know someone’s specific denomination to guess their position on school prayer—knowing their party is usually sufficient.

At the same time, these results push back against an overly reductionist view that religion is now entirely subsumed by partisanship. The findings indicate that even after controlling for party and general ideology, religious identity exerts residual effects. Notably, evangelical Protestants remain more supportive of school prayer than their partisan counterparts would predict. This suggests that the experience of belonging to an evangelical community—with its theological emphases on public prayer, witnessing one’s faith, and perhaps perceiving a secularizing threat—instills a propensity that goes above and beyond simply being a conservative Republican (A. E. Smith and Boas 2024). Likewise, among Democrats, those with evangelical or fervent religious backgrounds diverge from the party’s secular segment. Here emerges the persistence of a “religious left” or at least a religious Democratic contingent (often African Americans and Latinos) that, while loyal to the party on many issues, holds more traditional views on religion in the public square (McCarthy et al. 2019). This tension can occasionally surface in intra-party debates—for example, Black Democratic lawmakers have sometimes supported school prayer amendments or opposed measures seen as hostile to religion, putting them at odds with the predominantly secular liberal wing of the party. While the partisan divide explains much of the variance, these findings affirm that religious identity—especially evangelical affiliation—still matters in shaping public attitudes on prayer in schools.

How significant are these within-party differences in practice? On one hand, partisan sorting ensures that most high-salience policy debates—such as school prayer—are contested primarily between unified party blocs. The Republican Party generally supports teacher-led school prayer, with some lawmakers even proposing constitutional amendments to explicitly allow it, while the Democratic Party largely opposes such measures on First Amendment grounds (Saylor et al. 2022). As a result, internal dissenters—such as secular Republicans or devout Democrats—tend to be politically marginalized within their respective coalitions. While a significant percentage of mainstream Protestant Democrats in the sample still expressed support for teacher-led prayer, the party leadership and the official platform were largely opposed to such measures. These voters may deprioritize the issue or accept a compromise to maintain partisan alignment on other matters. Similarly, secular Republicans who may feel ambivalent about religious intrusions into public education nonetheless remain within the party due to alignment on tax policy, gun rights, or other core GOP issues (Schulman and Zelizer 2008). This dynamic reflects theories of coalition dominance, wherein party positions often reflect the preferences of ascendant internal factions. On cultural issues, the Republican coalition is strongly influenced by its evangelical wing, which exerts disproportionate control over the party line (Grossmann and Hopkins 2016). As a result, dissenters within both parties often have limited ability to reshape party agendas or public messaging.

On the other hand, residual religious cleavages within parties still carry important implications, particularly in shaping rhetorical strategies and coalition management. Among Democrats, awareness that a significant subset of their base—especially in districts with large Black or Latino populations—holds pro-religion views often leads candidates to frame messages that affirm the compatibility of faith and Democratic identity. For instance, President Obama frequently emphasized this theme, notably in his 2006 “Call to Renewal” address, where he asserted that “secularists are wrong when they ask believers to leave their religion at the door before entering the public square” (Arsi 2023).

Recent electoral evidence suggests that this religious divide inside the Democratic coalition is beginning to appear in vote choice. Pew Research Center’s validated-voter study of the 2024 presidential election shows that Donald Trump drew nearly even among Hispanic voters (a 3-point Democratic edge, down from 25 points in 2020) and more than doubled his share of the Black vote, from 8 percent to 15 percent (Pew 2025b). The same tables indicate that these gains were concentrated among voters who attend religious services at least monthly—including both Latino Protestants and Black Protestants. Taken together, the findings imply that the Democratic Party’s religious–secular cleavage, long visible in survey attitudes, is beginning to spill over into actual voting behavior, particularly among devout Black and Latino Democrats.

Similarly, Republican leaders occasionally temper explicitly sectarian language to avoid alienating secular-leaning conservatives or moderate suburban voters. Rather than directly advocating school prayer—a position that may unsettle some libertarian or pluralist factions—GOP elites increasingly invoke the more inclusive frame of “religious liberty,” which accommodates both evangelical priorities and broader constitutional values (Berg 2023). These patterns underscore how within-party religious diversity, even when muted in legislative outcomes, requires ongoing discursive balancing by party elites to maintain coalition coherence (Hilton 2021). Moreover, the persistence of within-party differences means that religious identity can still realign or reshape politics under certain conditions. If partisan polarization were ever to lessen on other dimensions (say, if cross-pressures on class or race emerged), these latent religious differences could become more salient again. One could envision scenarios, for example, where a charismatic religious Democrat could mobilize the devout Democrats and perhaps pull some religious moderates from the GOP, stressing common faith-based values, akin to Jimmy Carter’s appeal in 1976 or some local examples (Epter 2024). Conversely, a faction of the GOP that is more secular (some young Republicans, perhaps) might one day assert itself if the party’s electoral fortunes depend on moderating its religious conservatism. Thus, the religious identity effect is like a tectonic fault line beneath the parties: mostly the plates (parties) move in unison with the fault (religion) now, but stress can cause slips along the fault line.

This analysis specifically examines school prayer as a case study, but similar implications likely extend to other prominent issues at the intersection of religion and politics. For example, attitudes toward abortion, same-sex marriage, and the teaching of evolution versus creationism demonstrate a parallel pattern: substantial partisan divides reflecting religious sorting, alongside residual but increasingly rare internal party differences, such as small minorities of “pro-life Democrats” or “pro-choice Republicans,” whose positions frequently stem from personal religious convictions (Pew 2025a). In broader morality–policy domains, religious identity continues to shape individual consciences, occasionally overriding partisan directives (Budde et al. 2018).

However, it is notable how small these intra-party minorities have become—a striking manifestation of what scholars call “religious-political polarization” (Putnam and Campbell 2010). In previous eras, notably the 1970s, significant factions of traditionalist Catholics within the Democratic Party and secular or moderate Protestants within the Republican Party were common (Allitt 2019). Today, these groups have dramatically diminished. Indeed, the analysis of the Pew Religious Landscape Study (2023–2024) found secular Republicans or unaffiliated conservatives to be numerically insufficient for robust statistical analysis. Similarly, White evangelical Democrats are relatively rare outside the African American community. Thus, while complete alignment or sorting has not yet been reached, the degree of religious–partisan sorting observed today is extensive and highly consequential.

A salient instance of residual religious influence emerges when considering race in conjunction with religion. In the analysis, Black Democrats demonstrate higher support for school prayer than might be expected given their partisan affiliation, suggesting that racialized religious traditions, particularly the Black church, sustain a distinctive religious–political profile within the Democratic coalition. While the broader party platform opposes school prayer, a sizable subset of religious African American Democrats diverges on this issue, reflecting the difficulty of full ideological absorption. This reinforces the broader finding: even within partisan blocs, religious identity remains a powerful and sometimes independent force. While the core focus is on religious cleavages, the analysis underscores the importance of their entanglement with racial and ideological divides, which collectively deepen the intensity and durability of political polarization in contemporary America.

What do these findings mean for American democracy and political discourse? One interpretation is somewhat pessimistic: since partisan loyalties have absorbed religious divides, political debates about issues like school prayer may lack the cross-cutting dialogue they once had. When religious disagreements were less tightly aligned with party affiliation, there was more chance for bipartisan coalitions: for instance, a coalition of religious conservatives from both parties could team up on an issue, or secularists from both sides could oppose. Now, such coalitions are rare. Instead, debates become zero-sum partisan battles where each side invokes religion either as a rallying cry (on the right) or a threat (on the left). This intensification can contribute to a sense of cultural polarization and even the demonization of the other side as either “Godless” or “theocratic” (each side viewing the other through the worst religious lens). Indeed, scholars like Abramowitz (2018) and others note that affective polarization—dislike of the opposing party—is fueled by these stacked identity differences. An evangelical Republican not only disagrees with a secular Democrat on policy, but may see them as alien in values, and vice versa.

On a more optimistic note, the presence of intra-party ideological heterogeneity suggests that internal dissenters may act as a moderating force against the pull of extremism. For example, religious moderates and faith-based minorities within the Democratic Party—though numerically limited—have prompted efforts to temper overtly secularist rhetoric (Wadsworth 2014). Within the Republican Party, libertarian-leaning or religiously unaffiliated members have at times voiced discomfort with overt sectarianism, especially where it risks alienating younger, less religious voters (Keckler and Rozell 2015). These internal counterbalances can function as stabilizing agents, reminding party elites that neither party is entirely ideologically pure and that a measure of the “other” resides within their own coalition. While the practical influence of such actors may be limited, their existence introduces a layer of pluralism that tempers totalizing narratives and may offer a path—however narrow—toward greater political empathy.

It is also instructive to consider generational dynamics. Younger Americans are substantially more likely to identify as religiously unaffiliated and tend to lean Democratic (Hout and Fischer 2014). In contrast, the Republican coalition continues to draw heavily from older and more devout religious voters. As generational turnover progresses, the Democratic Party may become increasingly secularized—particularly if the influence of older African American Protestants wanes—potentially reducing even the residual religious variance observed today. Conversely, the Republican Party faces a strategic dilemma: how to maintain support among younger, less religious cohorts without alienating its existing evangelical base (Swartz 2012). While doubling down on Christian nationalist rhetoric may energize core supporters, it risks further distancing youth. The data reveal that Republicans under 30 were slightly less uniformly supportive of school prayer than their older counterparts, suggesting the early signs of generational divergence. If secular Republicans grow in number, internal tensions over cultural issues may become more pronounced. Alternatively, if generational sorting continues apace—where young religious conservatives remain in the GOP and young seculars remain in the Democratic fold—then polarization will be further entrenched, divided along both ideological and religious generational lines.

In conclusion, this analysis demonstrates that “who prays together” in America is increasingly “who votes together.” Partisan polarization has aligned itself with religious divides so extensively that knowing someone’s party nearly tells you their stance on public prayer in schools. This is a stark change from earlier eras when religion was a cross-cutting identity, creating Democratic evangelicals and Republican moderates in significant numbers (Perry 2022). Today’s parties are sorted: a largely religious right vs. a largely secular left, confirming the narrative of culture war polarization (J. J. Castle and Stepp 2021). Yet, by probing beneath these macro-alignments, the findings reveal the story is not one of total convergence of religion into partisanship. The endurance of intra-party religious effects demonstrates that religious identity remains a distinct layer of social identity. It can tug at individuals even when partisan currents push the other way. This nuance is important for scholars of political behavior, as it suggests that models treating partisanship as the be all and end all might miss subtle, but meaningful variations. It also matters for political strategists and civic leaders, who should recognize the remaining diversity within their camps.

Ultimately, while partisan polarization in the U.S. has absorbed most religious cleavages to the point that political fights often appear as fights between religious traditionalism and secular progressivism, it has not absorbed them completely. Residual effects of religious identity persist, serving as a reminder that religion for many Americans is not just another preference to be dictated by party, but a core personal identity that can inspire its own allegiance. The balance of power between these dueling loyalties—party and faith—will continue to shape the evolution of American political conflict in the years to come. As American society becomes more religiously polarized and the parties double down on identity-based appeals, understanding these nuanced dynamics will be crucial (Mounk 2023). This research contributes a detailed examination of one facet of that puzzle, and future research should explore others (e.g., how these patterns play out on different policy domains, or in electoral behavior such as split-ticket voting or primary contests).

6. Conclusions

This study set out to assess whether partisan polarization in the United States has fully subsumed religious cleavages or whether independent effects of religious identity still manifest within partisan groups. Drawing on recent nationally representative data from the Pew 2023–2024 Religious Landscape Study, the analysis focused on school prayer as a symbolic test case. The findings affirm that partisan polarization now absorbs most, though not all, religious divisions. Historically significant religious distinctions—such as those between evangelicals and mainline Protestants or between frequent and infrequent churchgoers—now largely correspond to partisan alignments (Woodberry and Smith 1998). In multivariate models, party affiliation emerged as the dominant predictor of support for teacher-led prayer, substantially diminishing the explanatory power of religious tradition. This confirms the broader pattern of partisan sorting along cultural lines: the Republican coalition encapsulates more devout and religiously traditional Americans, while Democrats increasingly represent secular or religiously liberal constituencies. In this sense, political identity has become a “proxy variable” for religious worldview. Nevertheless, religious identity has not entirely vanished from the political landscape. Even after accounting for partisanship, indicators such as evangelical affiliation and church attendance retain statistically and substantively significant effects, particularly within Democratic ranks. These intra-party cleavages underscore the persistence of religious diversity even within increasingly ideologically sorted parties.

At a theoretical level, this analysis advances scholarly understanding of identity politics by providing empirical insights into the formation of consolidated partisan mega-identities (Gorski and Perry 2022). It offers empirical evidence for the fusion of religion and party affiliation while also highlighting its limits. Not all identity components are perfectly aligned: cross-pressured individuals—those whose religious views conflict with their party’s dominant stance—continue to exist and matter. These findings suggest that these voters may adopt party-congruent positions publicly, but harbor private reservations, reflecting the continued relevance of classical theories of cognitive dissonance and cross-pressures (Dassonneville 2023). This residual ambivalence presents both a challenge and an opportunity for political strategists. For Democrats, appeals to secular values may fall flat among older or more religiously committed voters of color. For Republicans, overt religiosity may alienate more libertarian or less religious supporters. The partisan absorption of religion is thus not total: it leaves behind internal fissures that can reemerge in primaries, platform debates, or shifting electoral coalitions. More broadly, the persistence of these religious signals within partisan identities reinforces the utility of religion as an analytical lens in political behavior research.

Future research should further investigate how religion interacts with other salient identity domains—such as race, class, and gender—in shaping political attitudes and behaviors. The focus on school prayer, while theoretically justified and empirically fruitful, necessarily limits generalizability to other policy areas (e.g., reproductive rights, religious exemptions in law). Likewise, the analysis concentrated on Christian subpopulations due to data constraints, and the results are substantively identical when non-Christian respondents are retained (see Appendix A). As partisan coalitions harden around diverging religious worldviews, scholars must track how generational change, immigration, and issue realignment might further entrench or blur these divisions. Early evidence that highly religious Black and Latino Democrats are beginning to shift their vote choice reminds us that the party’s historic advantage among voters of color is not guaranteed and that religion may once again prove an engine of partisan realignment. Documenting whether such nascent movements endure will require panel studies that link validated ballots with granular measures of religiosity and social conservatism. Ultimately, partisan polarization remains the dominant organizing principle of American political conflict, but it has not rendered religion politically irrelevant. Instead, religion now operates both through and within partisanship, shaping values, coalitions, and contestation (Barton et al. 2021). Recognizing this dual structure is vital for understanding—and potentially bridging—the growing gap between civic and sacred values in 21st-century American politics.