Politics may make for strange bedfellows, but some matches seem downright inexplicable. Such may appear to be the case with evangelicals and Donald Trump. Trump's personal history—including multiple instances of fraud, tens of thousands of lies (e.g., Kessler, Reference Kessler2021), numerous cases of adultery, a legal finding that he committed sexual assault, caustic denigration of others, stoking of racial divisiveness, and a lack of concern with or understanding of Christian principles—arguably makes him the unlikeliest of presidential candidates for evangelicals to back (see also Djupe and Calfano, Reference Djupe, Calfano, Djupe and Claassen2018). As Dean and Altemeyer observe (Reference Dean and Altemeyer2020, 174 and 178), “Trump is the antithesis of most everything Christian ministers preach about the qualities of a good person and living a righteous life… (E)vangelicals should have spotted Donald Trump from the start as an enormously phony Christian.” Despite these seemingly salient aspects of Donald Trump's personal history, evangelicals provided Trump with a sizeable base of support in his 2016 and 2020 presidential campaigns (e.g., Newport, Reference Newport2020) and the 2024 Republican presidential primaries (e.g., Prude, Reference Prude2024).
By deepening our understanding of evangelical support for Donald Trump, insights about evangelicals' social and political motivations may be gained, resulting in greater clarity regarding the potential impact of evangelicals in U.S. politics. But what steps are needed to unpack why so many evangelicals have backed Donald Trump? We see at least two critical dimensions. One pertains to content, meaning the basis or bases on which evangelicals have come to favor Trump. In short, what substantive considerations have drawn evangelicals to support the former president? A second dimension involves intensity, which refers to whether evangelicals have supported Trump strongly or tepidly. In assessing evangelical support for Donald Trump, is that support best interpreted as enthusiastic or reluctant? Although implications of our approach for content will be noted, this study's central concern is the second dimension, intensity. Specifically, we consider the strength of evangelicals' support for Donald Trump near the mid-point of his presidency, and we use multiple measures and approaches to calibrate whether, in both absolute and relative terms, that support is best characterized as strong or moderate in intensity.
The approach we follow builds on that taken by Djupe (Reference Djupe2017). Selecting on self-reported 2016 Trump voters, Djupe's analysis of exit poll data shows no difference in Trump feeling thermometer ratings for evangelicals and nonevangelicals, nor any difference in the percentage of respondents in each group who voted for Trump despite assigning him a feeling thermometer rating below 50. Those findings suggest that evangelical support for Donald Trump in the 2016 election is more accurately characterized as enthusiastic rather than hesitant.
The analyses we report expand in several manners on Djupe's (Reference Djupe2017) initial assessment. First, rather than consider what Trump supporters felt at the moment of the 2016 election, we examine their opinions once Trump had been in office for over two years. We do so because information contexts change post-election, and because evaluations of a person as president may differ in form and favorability from evaluations of a person as a presidential candidate since candidate evaluation is an inherently comparative and party-driven process (Beasley and Joslyn, Reference Beasley and Joslyn2001; Dinas, Reference Dinas2014).
Second, recognizing the subtleties and possible multidimensionality of political support, we contrast evangelicals and nonevangelicals on six dimensions. This expansive approach helps ensure that no differences between evangelical and nonevangelical Trump supporters are overlooked. Third, two multivariate estimation strategies are used so that we can observe the possible effects of evangelicalism on Trump support while accounting for possible confounders. Specifically, we use OLS models to maximize the number of cases we can examine, and we follow with Mahalanobis distance matching to maximize the precision with which evangelicals and nonevangelicals are contrasted. Fourth, conditional specifications that incorporate religiosity and information exposure are added to test for possible variation within evangelicals in the intensity of their Trump support.
In the next section, we develop the case for the importance of assessing the intensity of evangelical support for Donald Trump. As part of this rationale, we explain why both intense and tepid support plausibly might be seen, and we detail scenarios in which variation in the intensity of Trump support may map to variation in the substantive bases on which Trump has been evaluated. Following this, the properties of our 2019 survey are discussed, along with the measures and procedures to be used in multivariate analyses. Results then are presented in several steps, moving from a multifaceted comparison of evangelicals and nonevangelicals to consideration of the possible moderating effects of religiosity and information exposure. The paper concludes with a discussion of the implications of our findings for future research seeking to refine our understanding of evangelical support for Donald Trump and assessments of evangelicals as political actors.
Perspectives on Evangelicals' Support for Donald Trump
From the start, Donald Trump's relationship with evangelicals has seemed perched on a foundation of uncertain strength. One perspective holds that evangelicals have backed Trump with caution and hesitancy. In 2016, for instance, it appeared that many evangelicals had concluded that while Donald Trump may have been a suboptimal and even disgusting candidate, he remained the best option among the likely winners (Scott et al., Reference Scott, Killough and Burke2016). The most prominent example of this viewpoint was voiced by Franklin Graham, who, in 2016, counseled his followers, “you may have to hold your nose and vote”—something that would not have been necessary were evangelicals drawn to Trump with unfettered zeal (qtd. in Law, Reference Law2016; see also Djupe, Reference Djupe2017). Further suggestions of hesitancy have followed, including claims that evangelicals struggled with the moral contradictions in evidence once the Trump presidency began (Keen, Reference Keen2018), signs that Trump's grip on evangelicals weakened following the January 6, 2021 assault on the Capitol (“Evangelicals are Divided”, 2021), the apparent movement of some evangelicals away from Donald Trump and toward Ron DeSantis in the run up to the 2024 Republican primaries (Thomson-DeVeaux and Cox, Reference Thomson-DeVeaux and Cox2023), Trump's ensuing claim that evangelical leaders were exhibiting “disloyalty” by not rushing to endorse his 2024 campaign (Holmes et al., Reference Holmes, Orr and Collins2023), and polling data showing that an increasing number of evangelicals might support Joe Biden over Donald Trump in the 2024 general election (Impelli, Reference Impelli2024).
The contrary perspective also has its advocates. The aforementioned analysis by Djupe (Reference Djupe2017) revealed no sign that evangelical Trump voters in the 2016 general election were in any way half-hearted. Prior to that, a hint of hesitation toward Trump may have existed in the 2016 primaries because evangelicals with higher levels of church attendance favored other candidates by a small margin in early contests (e.g., Layman, Reference Layman2016), but they quickly came around to support Trump at very high rates once he became the nominee. The oft-repeated statistic is that 81% of white evangelicals voted for Trump in the 2016 general election (Martinez and Smith, Reference Martinez and Smith2016), and the exit poll numbers have been corroborated with survey data (e.g., Burge, Reference Burge2017; Whitehead et al., Reference Whitehead, Perry and Baker2018; Margolis, Reference Margolis2020). Moving past the 2016 election context, subsequent reports found that evangelical support for Donald Trump remained vibrant despite the dozens of criminal indictments he faced (Williams, Reference Williams2023), that the commitment of evangelical voters had not changed from 2016 to 2020 to 2024 (Strong, Reference Strong2024), and that, if anything, by 2024 evangelicals had gone “all in” for Trump (Perry, Reference Perry2024).
Voting for Donald Trump should not be conflated with supporting him enthusiastically. Unfortunately, much of the commentary on whether evangelical support for Donald Trump is enthusiastic or reluctant in form reports impressionistic or anecdotal accounts. Building on Djupe's (Reference Djupe2017) direct test following the 2016 election, we develop a more systematic and multifaceted approach to addressing this question. With focus on the approximate midpoint of Trump's presidency, we seek to marshal as much evidence as possible regarding the intensity dimension of evangelical Trump support. Although our primary objective entails resolution of the intensity question, results also may speak indirectly to the viability of alternate accounts regarding content, or the substantive bases of the link between Donald Trump and his evangelical supporters. This is because some of the perspectives on why evangelicals side with Trump match better to either the enthusiastic or reluctant views of intensity. Thus, if clear evidence emerges regarding intensity, that evidence may help direct subsequent research on content.
If evangelicals offer Donald Trump only lukewarm support, one plausible explanation in terms of content is that evangelicals perceived transactional value in siding with Trump. On a personal level, Donald Trump's character and morality may be off-putting to evangelicals who favor both moral character and policy leadership consistent with their Christian values and principles. But, seen pragmatically, it might be better to side with a personally reprehensible politician if doing so brings wins on the policy front. This is what Franklin Graham meant by “hold your nose and vote for Trump.” Trump vowed to promote the evangelical policy agenda in exchange for their political support. For their part, joining that transactional framework may have been politically expedient for evangelicals, but they certainly did not have to like it. After all, no amount of Supreme Court Justices would erase the reality of Donald Trump the person. Policy-minded evangelicals reasonably may have voted for Trump and offered him minimal support, yet still felt coolly toward him and perceived him to be deficient on criteria such as honesty and morality. It follows that if our analyses show evangelical support for Donald Trump to be of weak intensity, that finding would bolster the viability of a transactional, policy-focused view of why evangelicals have sided with Trump.
What decision-making process would be consistent with a finding that Donald Trump's evangelical supporters, both en masse and across multiple dimensions, have hitched themselves to Trump with discernible enthusiasm? In this scenario, it would seem plausible that some force other than Christian values and principles motivates their behavior. Multiple often interconnected theses have been advanced. One with arguably the most traction is that Trump explicitly offered to restore Christians to a powerful position in the U.S. (Gorski, Reference Gorski2017; Whitehead et al., Reference Whitehead, Perry and Baker2018), and relatedly, that Trump himself provided strong guidance to evangelical followers who score high on right-wing authoritarianism (RWA – Dean and Altemeyer, Reference Dean and Altemeyer2020). Consistent with these perspectives, a Christian nationalist worldview continued to have a tight relationship with Trump support (Baker et al., Reference Baker, Perry and Whitehead2020) through to January 6 and beyond (Armaly et al., Reference Armaly, Buckley and Enders2022; Djupe et al., Reference Djupe, Lewis and Sokhey2023).
Christian nationalism is not the only dimension on which enthusiastic evangelical support for Trump might be expected. Guth (Reference Guth2019), for example, finds evidence that evangelicals are more likely to be populists, a descriptor often applied to Trump's politics (e.g., White, Reference White2016). Another throughline is the racial conservatism that tends to characterize white evangelicals, which was a dominant force explaining their 2020 vote in at least one analysis (Baker et al., Reference Baker, Perry and Whitehead2020).
Perhaps the most jaw-dropping suggestion is that evangelicals were so threatened by the shifts in politics and demographics in the United States that they were willing to change their fundamental values. Evidence aggregated by PRRI (2016) shows that evangelicals once were the most opposed to the idea that “an elected official who commits an immoral act in their private life can still behave ethically and fulfill their duties in their public and professional life.” Whereas only 30% of white evangelicals agreed with that statement in 2011, 72% agreed with it in 2016. An “evangelical crackup” with the GOP was clearly not in the making (Djupe and Claassen, Reference Djupe and Claassen2018).
The possible link between the intensity of evangelical Trump support and the substantive basis of that support can be explored by contrasting evangelicals and nonevangelicals. If principled evangelicals pursuing a policy agenda anchored in concern about values and morality backed Donald Trump reluctantly, then an intensity gap should be seen when levels of evangelical and nonevangelical Trump support are contrasted. Conversely, if evangelicals have sided with Trump due to factors such as a thirst for power rooted in Christian nationalism and a desire to slow the nation's changing demographics, then an intensity gap would be unlikely because Donald Trump has constituted a strong ally in pursuing those objectives and his moral and character deficiencies should not have muted enthusiasm about that alliance.
Two hybrid accounts also warrant consideration. First, seen in the political realm, the evangelical label might encompass multiple perspectives. According to this view, assessing four core religious beliefs—the so-called Bebbington (Reference Bebbington1989) Quadrilateral (e.g., Kidd, Reference Kidd2016; Keller, Reference Keller2017)—would enable analysts to differentiate between enthusiastic Trump support among nominal evangelicals and more tepid Trump support among true believers. Some prior research casts doubt on whether this distinction will be seen. For instance, Margolis (Reference Margolis2020) shows convincingly that nominal evangelicals are more likely to have supported Clinton or someone else in 2016. Although our survey lacks data on core beliefs, we will explore heterogeneity among evangelicals in broad form by assessing whether religiosity moderates evangelical Trump support.
A second hybrid account speaks to a possible information effect. It may be that all evangelicals potentially would see Donald Trump as personally repugnant, but those perceptions require awareness of problematic aspects of Trump's character. Evangelicals who tune out politics or do not look beyond slim information silos may have no basis on which to assess Donald Trump negatively. Given the widespread, multi-year discussion of Trump's personal failings, this scenario might seem unlikely, but a direct test is possible by factoring in information exposure. If many evangelicals warmly embrace Donald Trump only because they lack information that he is, as Dean and Altemeyer phrased it, “an enormously phony Christian” (2020, 178), then the intensity of Trump support among evangelicals should fall as information exposure rises.
The case developed here maps to a multi-step empirical strategy. First, examination of data from a period clear of the electoral context facilitates the assessment of views of Donald Trump the president rather than of Donald Trump the presidential candidate. Second, contrasts of views about Trump as expressed by evangelical and nonevangelical Trump supporters promise to provide direct evidence on the question of intensity and may yield indirect evidence on the matter of content. Third, conditional tests that incorporate information on religiosity and information exposure offer leverage on hybrid accounts that hold that some, but not all, evangelicals supported Trump reluctantly.
Data and Methods
All data are from an internet survey fielded in the United States in the period March 9–14, 2019. The survey was administered by YouGov and was completed by 2,500 respondents. The survey addressed topics related to political information and opinion and multiple items were included to measure facets of possible support for Donald Trump. The survey was conducted just past the midpoint of the Trump presidency. This places it after Charlottesville, the Gorsuch and Kavanaugh nominations and confirmations, the start of the Mueller investigation, the 2018 midterm elections, and the criminal convictions of Trump associates Michael Cohen, Michael Flynn, and Paul Manafort, but before the Mueller report was delivered, and well before the COVID pandemic, the two Trump impeachments, and the events related to the 2020 election.
Analyses center on Trump supporters. Two operationalizations are used. The first includes only those respondents who reported that they voted for Donald Trump in 2016 (N = 871). The advantage of this operationalization is that it enables us to gauge subsequent enthusiasm for Trump among his actual 2016 voters. A disadvantage is that it excludes Trump supporters who, for whatever reason, did not vote in 2016. Despite not voting in 2016, these individuals may have been politically engaged during the Trump presidency and may have planned to vote for him in 2020. To incorporate such respondents, a second operationalization includes 2016 Trump voters plus any respondents who did not vote in 2016 but who either awarded Donald Trump a higher feeling thermometer rating than Hillary Clinton or who indicated that they planned to vote for Trump in 2020. This operationalization increases the number of respondents identified as Trump supporters by just over 25% (N = 1,090). Critically, neither operationalization preordains that respondents identified as Trump supporters must be enthusiastic toward him. His 2016 and prospective 2020 voters may have perceived him to be the lesser of two evils, and he could have received a score as low as one (to Hillary Clinton's zero) on the feeling thermometer.
To produce an expansive perspective on whether evangelical and nonevangelical Trump backers differ in the enthusiasm of their support, six dependent variables are used. Three incorporate data from feeling thermometers. The first, Trump FT, is the 0–100 Trump measure. This provides a holistic take on Trump support. If evangelicals are cooler toward Donald Trump than are nonevangelicals, a difference in feeling thermometer scores will be observed. For 2016 Trump voters, the mean on this variable is 82.52 (s.d. = 22.78); for Trump supporters in our expanded operationalization that includes 2016 nonvoters, the mean is 79.23 (s.d. = 24.63).
The second dependent variable, Trump-Pence, represents the difference between the respondent's feeling thermometer scores for Donald Trump and Mike Pence. By 2016, Vice Presidential nominee Pence was strongly linked to evangelicalism both in his home state of IN (e.g., Sikich, Reference Sikich2016) and nationally (e.g., Boorstein, Reference Boorstein2016), and analyses suggest that Pence's reputation as a Christian conservative helped the Trump-Pence ticket weather challenges linked to Donald Trump's moral character, and especially the release of the Access Hollywood video (Merwin, Reference Merwin2023). This is a likely place to search for signs of hesitation in Trump support because, even for evangelicals committed to Donald Trump, the relative ratings of Trump and Pence should be expected to lean more toward Pence for evangelicals than for nonevangelicals. The means on this measure are 2.09 (s.d. = 18.24) for 2016 Trump voters and 3.15 (s.d. = 20.29) for all Trump supporters (positive numbers signal warmer feelings toward Trump).
The third dependent variable is used to address the possibility that, as political actors, evangelicals are motivated primarily by traditional conservative Republican principles. This variable captures the difference between warmth toward Trump and average warmth toward two prominent traditional conservative Republicans, Senators Ted Cruz and Mitch McConnell; specifically, Trump-Republicans = Trump FT – ((Cruz FT + McConnell FT)/2). Descriptive statistics for 2016 Trump voters are: mean = 21.01, s.d. = 20.59; the corresponding statistics for all Trump supporters are mean = 20.24, s.d. = 21.27. These marks reveal that Trump supporters are, on average, much more favorable to Trump than to other Republican leaders.
The remaining dependent variables draw on items other than feeling thermometers. Morality is a summative index that includes data from four questions tapping the extent to which respondents agree or disagree with allegations of wrongdoing against Donald Trump. Pro-Trump values indicate that respondents believe (1) the Trump administration was one of the least corrupt in U.S. history, (2) the Mueller investigation was a witch hunt that produced no evidence of wrongdoing on the part of Trump's associates, (3) Donald Trump does not lie a lot, and (4) Trump probably has not committed several serious crimes. Each item is coded 0 (strongly anti-Trump) to 4 (strongly pro-Trump). For 2016 Trump voters, the 0 to 16 index is highly reliable (Cronbach's α = 0.79); descriptive statistics are: mean = 11.56, s.d. = 3.40. Including Trump supporters who did not vote in 2016, α = 0.80, mean = 11.13, and s.d. = 3.55. On the three feeling thermometer variables, it is possible that evangelical respondents pleased with Trump actions such as the Gorsuch and Kavanaugh nominations used the feeling thermometers as opportunities to express their support and gratitude rather than actual feelings of warmth. The morality scale is much less susceptible to such shifting in interpretation because the four items address very specific topics and behaviors of direct relevance to Donald Trump's character. Hence, this variable offers a more liberal test of the possibility that evangelicals were reluctant Trump supporters.
The fifth dependent variable, True-False, includes data from four items on which respondents were asked to gauge whether statements made by Donald Trump were true or false. In actuality, two of the statements were true.Footnote 1 Due to Trump's history of struggles with veracity, wary evangelicals concerned about Donald Trump's personal history should be highly attentive to the possibility that some of his claims were false. Therefore, the sum of how many times respondents answered “true” should be lower for evangelicals than for nonevangelicals. Descriptive statistics are mean = 2.94 and s.d. = 1.08 for Trump voters and mean = 2.87 and s.d. = 1.13 for all Trump supporters.
The final dependent variable pertains not to Donald Trump, but to respondents' views of their fellow Trump supporters. By the time the survey was fielded, respondents had the opportunity to know that Trump-supporting white supremacists and neo-Nazis had marched at Charlottesville and that several Trump advisers had been convicted of crimes. In this context, seven-point scales captured the extent to which respondents saw “most Trump supporters” as honest or dishonest, hard working or lazy, intelligent or unintelligent, and patriotic or unpatriotic. If the “hold your nose and vote for Trump” mentality serves partly as a means for evangelicals to differentiate themselves from Trump supporters who do not share their principles, then evangelicals should produce lower scores on the 0 to 24 summative scale. For 2016 Trump voters, Cronbach's α = 0.62, mean = 18.20, and s.d. = 4.31. For all Trump supporters, α = 0.67, mean = 17.64, s.d. = 4.63).
The survey includes the Pew religious identification items. We code respondents as being Evangelicals (1 = evangelical, 0 = nonevangelical) if they indicate that they consider themselves to be born again and their religion is either Protestant or “something else.” Nonevangelicals are respondents who do not consider themselves to be born again, or who are born again but who specify a different religion.Footnote 2 It is important to note a continued debate about using a denominational compared to an identification approach to classifying evangelicals (e.g., Burge and Lewis, Reference Burge and Lewis2018; Smith et al., Reference Smith, Sciupac, Gecewicz and Hackett2018; Smidt, Reference Smidt2019); we use the available identity-based classification strategy (limited by religious group) without the ability to comment on the broader debate. Among all respondents, 556 (22.2%) are evangelicals. Among 2016 Trump voters, 302 (34.7%) are evangelicals. For all Trump supporters, 362 (33.2%) are evangelicals.Footnote 3 Although some research on Trump support among evangelicals examines only white evangelicals, we do not impose this restriction because we select cases on the basis of Trump support, meaning nonwhite evangelicals who voted against Trump are excluded. To determine whether the inclusion of Trump-supporting nonwhite evangelicals alters results, core tests were re-run with only white evangelicals included (see Appendix A), but all results remained consistent with those reported below.
At question is whether Donald Trump's evangelical supporters view him more tepidly than do his nonevangelical supporters. In simplest form, testing requires only that we calculate mean differences for the two groups. That strategy risks conflating evangelical status with other variables. For instance, compared with nonevangelicals, evangelicals may be more conservative and they may view religion as more important in their lives, meaning failure to account for such factors could cause us to attribute effects to evangelicalism that actually trace to its correlates. To address this, all effects will be estimated in two manners. First, multivariate OLS regression models will be estimated for each dependent variable. Although the evangelical indicator measure is our chief independent variable, we control for multiple possible confounders, including political ideology, a 0–8 index of religiosity that incorporates information on frequency of church attendance and importance of religion (for 2016 Trump voters, mean = 4.54, and s.d. = 2.55; for all Trump supporters, mean = 4.39, and s.d. = 2.55), education, living in the South, race, and gender. Compared with nonevangelicals, evangelicals on our survey are more conservative, have lower education levels, are more likely to live in the South, and have higher religiosity. Because these are all plausible predictors of Trump support, we include them to avoid omitted-variable bias.
Our second estimation strategy is matching. We used optimal pair with exact matching on political ideology and religiosity. We attempted additional matching specifications, including nearest neighbor, full matching, and optimal without exact matching. However, these led to an inadequate balance of covariates.Footnote 4 Optimal pair with exact matching was estimated using Mahalanobis distance and all standardized mean differences (including squares and interactions) were not greater than 20% suggesting an adequate balance of covariates (Rosenbaum and Rubin, Reference Rosenbaum and Rubin1985). We fit linear models for each dependent variable and computed the marginal treatment effect on the treated (i.e., evangelicals) by running G-computations with robust standard errors clustered on the matched pairs. G-computation works by first fitting a linear regression for our outcomes given the treatment, covariates, and interactions between treatment and covariates. We include these interactions because of the remaining covariate imbalance and potential heterogeneity in the treatment. The model fit is then used to predict average outcomes under treatment and no treatment. The difference in point estimates is identical to the average treatment on the treated (Wang et al., Reference Wang, Nianogo and Arah2017).
Analyses proceed in three stages. First, OLS and matching tests will be conducted for the first group of Trump supporters – those who voted for Donald Trump in 2016. Second, tests are repeated for the expanded second group that includes both 2016 Trump voters and 2016 nonvoters who either planned to vote for Trump in 2020 or who awarded Trump a higher feeling thermometer score than they provided for Hillary Clinton.
In the third step, conditional tests are added. In the first of these, a second version of OLS models for both groups of Trump supporters will be estimated, this time including an evangelical × religiosity interaction. These models provide a coarse test of whether nominal evangelicals (individuals who identify as evangelical but neither see religion as highly important in their lives nor attend religious services with great frequency) were more supportive of Donald Trump than evangelicals for whom religion held a more central position. The second conditional tests are used to explore whether information differences among evangelicals lead to within-group variation in Trump support. Here, the logic holds that evangelicals only should have been reluctant to support Trump if they were aware of his personal failings, and low-information evangelicals may have lacked that awareness.Footnote 5 Drawing on research that factual knowledge is the strongest predictor of news acquisition (Price and Zaller, Reference Price and Zaller1993), we used data from a ten-item battery of awareness of nonpolitical current events (for 2016 Trump voters, mean = 7.40, s.d. = 1.78; for all Trump supporters, mean = 7.16, mean = 1.93). Our expectation is that respondents with wide knowledge of current events should be more likely to have heard about negative aspects of Donald Trump's personal history. Tests substituting a civics knowledge measure for current knowledge are reported in Appendix A; no noteworthy differences are observed versus the findings reported below.
Results
To test whether evangelical supporters of Donald Trump side with Trump more reluctantly than do Trump's nonevangelical backers, contrasts between the two groups are examined on six dependent variables, with each contrast investigated via both OLS and matching. Support for reluctance among evangelicals would be found if statistically significant negative effects are seen for evangelicalism, especially if those effects exhibit consistency across the multiple dependent variables and estimators.
Initial tests center on only those respondents who reported having voted for Donald Trump in 2016. The results are shown in Figure 1. Plots are point estimates of scores on the dependent variable with 95% confidence intervals. The first two estimates in each panel (those with squares) are derived from OLS models, whereas the third and fourth estimates (those with circles) are derived from matching. For each pair of estimates, the first depicts the point estimate and confidence interval for nonevangelicals (“NE”) and the second shows the outcome for evangelicals (“E”). The plots concisely summarize the key results from twelve multivariate models (coefficient estimates for OLS models are reported in Appendix A and matching results are reported in Appendix B). The findings are mostly inconsistent with the possibility that evangelicals offered Donald Trump more reluctant support than did nonevangelicals. None of the OLS tests produces a statistically significant negative effect. For the matching results, none of the contrasts between average marginal predicted outcomes for evangelicals and nonevangelicals is significant, although the ATT (average treatment effect on the treated) narrowly reaches the p < 0.05 level for the Trump-Pence measure, with evangelicals' relative feeling thermometer scorings of Trump and Pence favoring Pence by four points versus the ratings provided by nonevangelicals.Footnote 6 The effects also are substantively negligible. For instance, when the evangelical-nonevangelical gap is viewed relative to each variable's standard deviation, the very largest substantive effect is a mere 0.21 s.d. for Trump-Pence.
Figure 1. The intensity of Trump support among 2016 evangelical and nonevangelical Trump voters.
Note: Plots depict average marginal predicted outcomes with 95% confidence intervals. All contrasts are statistically insignificant. The plots are derived from coefficient estimates in twelve multivariate models; coefficient estimates for full models are reported in Appendix A, Tables A-13 and A-17; estimates for the matching models are reported in Appendix B, Table B-13 and B-14. To increase visual clarity, vertical axes are constrained to approximately one-half s.d. above and below the observed mean for each dependent variable. F-NE, full (OLS) model, nonevangelicals; F-E, full model, evangelicals; M-NE, matching, nonevangelicals; M-E, matching, evangelicals.
The point estimates for evangelicals also reveal an absence of hesitation. Evangelical Trump voters awarded Donald Trump an average feeling thermometer score of 82.5, and only 5.6% gave Trump feeling thermometer marks below 50. In contrast, 10.0% of nonevangelical Trump voters scored Trump below 50 on the feeling thermometer. Similarly, 94.7% of evangelical Trump voters approved of the job he was doing as president midway through his term, providing scant evidence of buyer's remorse. Evangelical Trump voters rated Trump nearly even with Mike Pence and twenty-one points above Ted Cruz and Mitch McConnell; they scored Trump well above the midpoint on the morality scale; on average they rated 2.99 of his four factual statements as true, even though two of Trump's claims were false; and they viewed their fellow Trump supporters as being honest, intelligent, hard-working, and patriotic. Collectively, the results reveal no sign that evangelicals were doubtful or unenthusiastic in their support of Donald Trump.
Figure 2 depicts results obtained when the twelve tests are repeated, this time in models that include all Trump supporters instead of only 2016 Trump voters. Although the number of cases increased by slightly more than 25%, the findings are nearly identical to those in Figure 1. No test yielded a statistically significant difference between evangelicals and nonevangelicals, and all substantive differences are negligible. Absolute effects among evangelicals again show favorable views of Donald Trump, including high average feeling thermometer scores, similar marks compared with Mike Pence, high ratings compared with other Republicans, and positive assessments of Trump's morality, honesty, and of the personal qualities of other Trump supporters.
Figure 2. The intensity of Trump support among evangelical and nonevangelical Trump supporters.
Note: Plots depict average marginal predicted outcomes with 95% confidence intervals. All contrasts are statistically insignificant. The plots are derived from coefficient estimates in twelve multivariate models; coefficient estimates for full models are reported in Appendix A, Tables A-21 and A-25; estimates for the matching models are reported in Appendix B, Table B-15 and B-16 To increase visual clarity, vertical axes are constrained to approximately one-half s.d. above and below the observed mean for each dependent variable. F-NE, full (OLS) model, nonevangelicals; F-E, full model, evangelicals; M-NE, matching, nonevangelicals; M-E, matching, evangelicals. Trump supporters include self-identified 2016 Trump voters plus 2016 nonvoters who indicated either that they planned to vote for Trump in 2020 or that their feeling thermometer rating for Donald Trump was higher than their rating for Hillary Clinton.
Our final tests examine possible heterogeneity among evangelical Trump supporters. The first conditional effects center on religiosity, which we use to differentiate nominal evangelicals from true believers. For these within-group tests, the possibility of reluctant Trump support among some evangelicals would be demonstrated if views of Trump are least favorable for those evangelicals who exhibit the highest levels of religiosity. We examine this possibility in a series of OLS models that add an evangelical × religiosity interaction. Each model is run twice, once for 2016 Trump voters and a second time including all Trump supporters.
Once again, models are estimated for 2016 Trump voters and for the expanded set of Trump supporters that includes 2016 nonvoters. Results are depicted in Figures 3 and 4. In these figures, the horizontal axis represents variation in religiosity, and separate estimates and confidence intervals are derived for evangelicals (solid lines) and nonevangelicals (dashed lines).
Figure 3. The intensity of Trump support with religiosity interaction, 2016 Trump voters.
Note: Plots depict fitted lines with 95% confidence bands. The plots are derived from coefficient estimates in six multivariate models; coefficient estimates for full models are reported in Appendix A, Tables A-14 and A-18; Plots are for 2016 Trump voters. Rug plots on the horizontal axis are values of religiosity. Along the vertical axis are values of the dependent variable. Models include the full 0 to 8 range for religiosity, but the plots include only 3 to 8 because no evangelicals recorded religiosity levels lower than 3.
Figure 4. The intensity of Trump support with religiosity interaction, all Trump supporters.
Note: Plots depict fitted lines with 95% confidence bands. The plots are derived from coefficient estimates in six multivariate models; coefficient estimates for full models are reported in Appendix A, Tables A-22 and A-26; Plots are for all Trump supporters. Rug plots on the horizontal axis are values of religiosity. Along the vertical axis are values of the dependent variable.
Results in Figures 3 and 4 provide modest evidence that some evangelical Trump supporters are reluctant. The patterns are largely consistent across the twelve sets of estimates.
Among evangelicals, some within-group variation across religiosity is observed, with high-religiosity evangelicals exhibiting somewhat softer support than their low-religiosity counterparts. This is seen in the negative slopes of the solid lines. All twelve solid lines in the two figures slope downward from left to right, indicating that, within evangelicals, high-religiosity evangelicals were the most lukewarm toward Donald Trump. For Trump FT in Figure 3, for instance, the estimated Trump feeling thermometer value decreases from 88.4 to 79.7 as religiosity rises from its lowest to its highest value. In both Figures 3 and 4, the within-evangelical effects reach statistical significance for the three feeling thermometer variables but not for morality, the true-false honesty measure, or for views of other Trump supporters.
At first glance, the within-evangelical results for the feeling thermometer variables seem consistent with a modified reluctant supporter hypothesis. However, closer inspection casts doubt on that initial assessment. First, significant effects do not emerge for assessments of Trump's morality or honesty, or for respondents' views of other Trump supporters, variables for which reluctance among high-religiosity evangelicals should have been especially likely. Second, the feeling thermometer point estimates for high religiosity evangelicals do not evince reluctance. Even high-religiosity evangelicals awarded Trump an average feeling thermometer mark near 80 and gave him a double-digit bump relative to McConnell and Cruz. Third, in no instance is the point estimate for high-religiosity evangelicals significantly lower than the point estimate for either high- or low-religiosity nonevangelicals. Hence, compared with nonevangelicals, we cannot conclude that high religiosity evangelicals exhibited reluctance.
To the extent that there is an outlier in Figures 3 and 4, it is low-religiosity evangelicals; all groups exhibit strong support for Donald Trump, but low-religiosity evangelicals are the most enthusiastic. In ten of the twelve panels—all but the two sets of estimates for the Trump-Pence comparison—the greatest favorability toward Trump is seen in the upper left portion, where the high marks for the solid lines reveal especially intense Trump support among low-religiosity evangelicals. Although these effects signify within-evangelical variation, they do not show reluctant Trump support. Instead, evangelicals report either strong support for Donald Trump (high religiosity) or very strong support (low religiosity).
Our final tests address whether an information effect produces within-group differences among evangelicals. If some evangelicals support Donald Trump only because they are unaware of his personal deficiencies, then high-information evangelicals—measured here using data on current events knowledge—should voice more negative views of Trump than low-information evangelicals. Estimates for 2016 Trump voters and all Trump supporters are reported, respectively, in Figures 5 and 6. As with the religiosity tests, within-evangelical differences will be established via the slopes of the solid lines, with sharper downward slopes, moving from left to right, signifying that greater information awareness among evangelicals corresponded with less enthusiasm toward Trump.
Figure 5. The intensity of Trump support with current events interaction, 2016 Trump voters.
Note: Plots depict fitted lines with 95% confidence bands. The plots are derived from coefficient estimates in twelve multivariate models; coefficient estimates for full models are reported in Appendix A, Tables A-16 and A-20; Plots are for 2016 Trump voters. Rug plots on the horizontal axis are values of current events knowledge. Along the vertical axis are values of the dependent variable. All values of current events knowledge are included in the statistical models, but the horizontal axis in Figures 5 and 6 is constrained to range from 4 to 10. This is because the current events items are three-category multiple-choice measures, meaning very few respondents received scores less than 4.
Figure 6. The intensity of Trump support with current events interaction, all Trump supporters.
Note: Plots depict fitted lines with 95% confidence bands. The plots are derived from coefficient estimates in twelve multivariate models; coefficient estimates for full models are reported in Appendix A, Tables A-24 and A-28; Plots are for all Trump supporters. Rug plots on the horizontal axis are values of current events knowledge. Along the vertical axis are values of the dependent variable.
The results in Figures 5 and 6 provide no evidence that evangelical support for Donald Trump hinged on a dearth of information. First, across twelve tests there are no instances in which Trump support among evangelicals grows significantly more tepid as current events knowledge rises. Second, the substantive changes as a function of current knowledge are consistently negligible. Third, the effects are directionally inconsistent, with five of the twelve slopes for evangelicals rising, not falling, as current events awareness heightens.
Conclusion
Despite a multifaceted effort to identify evidence that evangelicals were hesitant in their support for Donald Trump, no findings consistent with that expectation were observed. To the contrary, when viewed in absolute terms, evangelicals' assessments of Trump midway through his presidency were quite fervent. Moreover, relative to Trump's nonevangelical supporters, there was no instance across twenty-four tests in which evangelicals were found to be significantly more tepid toward Trump than were their nonevangelical counterparts. Conditional tests focused on religiosity also produced no evidence of reluctant evangelicals, although modest evidence emerged that low-religiosity evangelicals were particularly enthusiastic about Trump. Conditional tests incorporating current events awareness generated no evidence consistent with the possibility that evangelical support for Donald Trump stemmed from obliviousness regarding Trump's personal history.
On the fundamental matter of intensity, nothing in this study's results corroborates the possibility that evangelicals were reluctant Trump supporters. At the midpoint of Donald Trump's presidency, no hints of hesitation were seen among evangelicals. As with when they voted in 2016 (Djupe, Reference Djupe2017), evangelicals in 2019 saw no need to hold their nose when they expressed warmth for Donald Trump, when they awarded him high marks on morality and honesty, and when they commended the character of other Trump supporters.
A counterargument to our findings would be that perhaps Trump-supporting evangelicals on our survey felt personal disdain for Donald Trump but chose not to express that unfavorable view because they desired to use their survey responses as a means to boost a president whom they supported politically. In short, evangelicals actually were reluctant Trump supporters, but they opted to keep that reality to themselves. We do not see a strong case for this position. First, if evangelicals can exaggerate their Trump support for expressive purposes, so can nonevangelicals. If observed opinion is a function of true opinion plus an expressive bonus, then we still should have been able to detect reluctance among evangelicals by comparing them with nonevangelicals. Across twenty-four tests summarized in Figures 1 and 2, there is only a single instance—the Trump-Pence matching contrast for 2016 Trump voters—in which the difference in Trump support between evangelicals and nonevangelicals even hints at statistical significance. Second, we used six different dependent variables partly to hedge against expressive effects. Evangelicals who appreciated matters such as the Gorsuch and Kavanaugh nominations may have inflated their Trump ratings on the feeling thermometers, but it is less plausible that they would have deliberately misreported Donald Trump as being highly honest and moral if they believed otherwise. Third, with all empirical evidence indicating the absence of an evangelical-nonevangelical gap, the burden of proof shifts to proponents of the reluctant evangelical thesis to demonstrate that these null results are the product of expressive behavior.
We have suggested that evangelical support for Donald Trump can be assessed on the dimensions of intensity and content. On intensity, we see a dimension without group-based variation. At the midpoint of his presidency, all Trump supporters, including both evangelicals and nonevangelicals, assessed him quite positively. On content, or the basis on which evangelicals are drawn to Donald Trump, matters are more complicated. This is primarily because the evidence reported here is mostly indirect and circumstantial.
One account we have considered holds that evangelical support for Donald Trump reflects a transactional policy-focused mindset in which evangelicals begrudgingly backed Trump because, even while viewing Donald Trump the person as morally offensive, evangelicals recognized his empowerment served their policy goals. In our view, this study's findings are inconsistent with this account. Evangelicals awarded Donald Trump high marks both in absolute terms and relative to nonevangelicals. Further, they did so on metrics well removed from policy, such as ratings of Trump's morality and honesty, and assessments of his supporters. Evangelicals who saw Donald Trump as pragmatically useful but morally repugnant would not have found Trump to be honest or his administration to be squeaky clean.
Two conditional accounts we tested also failed to yield evidence that evangelicals were hesitant to back Donald Trump. First, it is not the case that high-religiosity evangelicals gave Trump poor marks. Low-religiosity evangelicals were the most enthusiastic Trump supporters we identified, but high-religiosity evangelicals equaled nonevangelicals in their favorability toward Trump. Second, we also can rule out an information effect. Collectively, the evangelicals in our study had high levels of awareness of current events. More critically, variation in awareness levels did not correspond with variation in the intensity of Trump support. These results are starkly inconsistent with the possibility that evangelicals backed Trump only because they were unaware of his problematic personal history.
Where this leaves matters, largely by default, is that evangelicals most likely have fervently latched themselves to Donald Trump because they feel an affinity with him that is rooted in something other than Christian values and principles. In this scenario, current null results would mean that evangelicals prioritized winning. Despite his personal flaws, it was good to have Donald Trump on their side. After all, Tony Perkins of the Family Research Council quipped, “You only have two cheeks. Look, Christianity is not all about being a welcome mat which people can just stomp their feet on” (Stanton, Reference Stanton2018). This interpretation provides indirect support for the Christian nationalist/RWA perspectives noted earlier. For evangelicals who were aware of Trump's troubling behaviors yet awarded him high marks anyhow, their prioritization of factors other than moral character seems self-evident: their support was in no way half-hearted, and they showed no signs of being wracked by either doubt or guilt. In fact, Djupe and Burge (Reference Djupe and Burge2018) reported that very few voters had regrets about their 2016 votes, especially among frequent church attendees. Our data do not speak to whether those other factors incorporate nationalist and authoritarian sentiment, but those accounts retain a viability that the reluctant Trump supporter hypothesis can no longer claim.
Looking ahead, our results suggest that a future candidate's personal limitations likely will not be viewed as disqualifying by evangelical voters provided that the candidate's messaging aligns with evangelicals' quest for political power. Christian commentator Dana Loesch effectively demonstrated this point in her vociferous defense of Herschel Walker, a 2022 GA Republican senate nominee with a personal history replete with glaring instances of dishonesty, hypocrisy, and alleged domestic violence. Rationalizing her support for Walker, Loesch said “I don't care if Herschel Walker paid to abort endangered baby eagles. I want control of the Senate” (Rubin, Reference Rubin2022). This circumstance gives rise to a broader question, one that encompasses all voters rather than just evangelicals: in contemporary American politics, is there anything in a candidate's personal history that would constitute an absolute deal-breaker? Heading into the 2024 presidential election campaign, 21% of survey respondents who indicated they intend to vote for Donald Trump answered in the affirmative when asked whether they think Donald Trump has or has not committed any serious federal crimes (“Cross-Tabs”, 2024). At least for these voters, the answer to our question appears to be no, there are no absolute deal-breakers.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1755048324000142
Acknowledgments
This research was supported by the Summer Research Opportunities Program (SROP) and the Benson Fund, both at the University of Illinois
Alondra S. Pagán Márquez is a recent graduate of the University of Puerto Rico, Río Piedras, and, as of August 2024, a graduate student in the Department of Political Science at the University of Illinois.
Paul A. Djupe is the Director of the Data for Political Research program at Denison University and the academic editor of the Religious Engagement in Democratic Politics series at Temple University Press.
Matthew Mettler is a doctoral student in the Department of Political Science at the University of Illinois.
Jeffery J. Mondak is the James M. Benson Chair in Public Issues and Civic Leadership in the Department of Political Science at the University of Illinois.
Open access
