The Denial Thesis: Checking a Popular Argument About Trump's Voters

A particular argument about Donald Trump and his voters circulates widely, usually in the form of a long social-media post. It runs roughly like this. Trump has no religious conviction, conservative Christians backed him anyway, and that willingness to overlook what they once condemned was the leading edge of a broader turn away from fact itself. Trumpism, on this reading, is organised denial, and its followers are motivated less by wanting anything for themselves than by wanting their opponents to suffer.

Parts of that are measurable. Parts of it are not measurable by anyone, because they assert what is going on inside millions of people's heads. This piece separates the two and rates the first kind.

The short version is that most of the individual facts survive, and the causal story attached to them does not. On the two claims where the data can actually test a sequence of events, the sequence runs the other way.

What the record shows about his religious knowledge

The evidence here is real and it is on the record. At Liberty University in January 2016 Trump cited "Two Corinthians, 3:17" rather than Second Corinthians, and reporters in the room from four outlets independently described students laughing.^[15] Asked in August 2015 to name a favourite Bible verse, he declined, saying it was personal. Asked again the following April, he offered "an eye for an eye."

One complication belongs in any fair account. Two days after the Liberty speech, Trump said the reference had been written out for him by Tony Perkins of the Family Research Council. Perkins confirmed it on the record, saying he had written the scripture reference himself and was "guilty as charged." Perkins also added that the episode showed Trump was not familiar with the Bible.^[16] The single most-cited proof of his biblical illiteracy is partly an artifact of someone else's notecard, and the man who wrote the card still reached the same conclusion.

Taken together the pattern of unfamiliarity is well documented, which is why this rates Mostly True rather than True. The individual anecdotes are weaker than their reputation.

What the record cannot show

The stronger version of the claim asserts an absence of conviction and ignorance of the core tenets of any faith. That reaches past anything the evidence can establish, and the counter-evidence is more substantial than the argument allows.

First Presbyterian Church in Jamaica, Queens confirmed to CNN that Trump attended Sunday school and was confirmed there. Marble Collegiate Church said his parents were active members for years and that one of his children was baptised there.^[17] Those statements come from the churches, not from him. In the same 2015 answer that produced his best-known religious gaffe, he volunteered an unprompted and specific recollection of Norman Vincent Peale's preaching, describing how Peale brought modern situations into a sermon.^[13] In 2020 he told Religion News Service that he had been confirmed Presbyterian as a child and now considered himself a non-denominational Christian.^[18]

None of that establishes sincerity either. Interior conviction is not a thing a survey or a transcript can reach. A claim about it is not false, it is unrateable, and it rates Unproven on those grounds.

A quote that gets hardened in the retelling

The most frequently repeated line in this argument is that Trump said he has never asked God for forgiveness. He did not say that. At the Family Leadership Summit in Ames, Iowa on July 18, 2015, Frank Luntz asked him twice. The second time, Trump answered: "I am not sure I have, I just go and try to do a better job from there. I don't think so."^[13][14]

A hedge became a denial somewhere in the retelling. The same answer produced the widely quoted remark about taking communion, which is nearly always cut in half. The full sentence is that when he drinks his little wine and has his little cracker, "I guess that is a form of asking for forgiveness. And I do that as often as possible, because I feel cleansed." Printed without its second half it reads as a dismissal of the sacrament. Printed whole it reads as a thin but sincere answer to the question he had been asked.

A fact-check that applies one standard has to correct this one too, so it rates Misleading.

The alignment itself is real

On the central empirical claim the argument is simply right, and every instrument agrees on the magnitude. The instruments disagree on the exact figure, which matters only if someone wants to claim a trend between cycles.

Pew's validated-voter series matches survey respondents against state turnout records rather than interviewing people at polling places.^[2][3] The network exit poll is a different instrument with different error.^[4][5][6] Anyone reporting a rise or fall between cycles has to name which one they are using, and the 2020 Pew figure was itself revised from 84% to 83% under updated weighting. The claim of overwhelming support rates True on any of them.

The argument is also right that evangelical voters were not under any illusion about him. Pew found in January 2016 that 44% of white evangelical Republicans viewed Trump as not too or not at all religious, and that 61% nonetheless thought he understood the needs of people like them, against 24% who said so of Hillary Clinton.^[1] The transactional reading of that relationship is supported by the people in it.

They were not his base when they had a choice

Where the argument slips is in treating this bloc as Trump's foundation from the beginning. During the primaries he lost white evangelicals outright in at least nine states, including Texas by 25 points and Wisconsin by 22. Across the twenty contests with exit polls up to late March 2016, Geoffrey Layman calculated that Trump averaged 36% of the white born-again vote, a plurality in twelve states, and slightly below the 38% he was drawing from all other Republican voters.^[11]

Evangelicals were marginally his weakest group relative to the rest of the party at the point when alternatives still existed. The frequently repeated refinement, that only the non-churchgoing ones backed him, is also too clean. Ted Cruz won the most devout slice, those attending more than weekly, while Trump won every other attendance category including regular weekly attenders.^[12] By July 2016 the gap had closed entirely, with Pew finding regular attenders just as supportive as infrequent ones, three-quarters in both groups.^[1]

Consolidation arrived with the general election and it was substantially about the other candidate. Among white evangelical Republicans who had wanted someone other than Trump nominated, 93% said by late June that they would vote for him. In the same survey, 55% said they were dissatisfied with the choice, and those describing their vote as mainly against Clinton outnumbered those describing it as mainly for Trump by 45% to 30%.^[1] The claim that they were his base from the start rates False.

The morality reversal, and the problem with its date

Here is the statistic the whole argument rests on, and it is real. PRRI asked whether an elected official who commits an immoral act in their personal life can still behave ethically in public office. In June 2011, 30% of white evangelical Protestants said yes. In October 2016, 72% did.^[7][8] Few attitudes move 42 points in five years.

Two things break the clean reading of it.

The first is the date. PRRI fielded that question from October 12 to 17, 2016. Evangelical consolidation behind Trump was already complete by late June, at 78%, and the fielding happened in the week after the Access Hollywood tape.^[1] The number therefore measures an attitude that arrived alongside or after the political decision, which means it cannot demonstrate a standard abandoned in order to make that decision. There is no reading of the question between June 2011 and October 2016, so no measurement exists that could settle the direction. The sequence the data supports is that they backed him and their stated standards moved to match.

The second is that the shift was not an evangelical story at all.

Democrats moved too, from 49% in 2011 to 61% by October 2016. PRRI's own conclusion was that no party is immune from fluctuation on the question.^[10] Evangelicals made the largest move and started from the lowest base, which is a real finding, but they were riding a wave that crossed party and denomination. Reading one column of that table as distinctive hypocrisy misreads the table.

Two cautions on the statistic itself. The 2016 white evangelical subsample was 180 people, and PRRI never published the 2011 subsample size at all, so one half of the headline shift carries no margin of error. Survey mode also changed at every reading, from landline-only in 2011 to a mostly online panel in 2020.^[9] The move is large enough to survive all of that. It is not a clean like-for-like series. The claim rates Mixed: the number is real, the causal story around it is not supported.

Did facts become irrelevant under Trump?

This is the argument's broadest claim and the one the data handles worst. Pew has tracked confidence that scientists act in the public interest for a decade.

Confidence rose through Trump's first term to the highest point in the series, fell during the pandemic, and has partly recovered.^[19] The decline is shaped like COVID, not like Trump.

Climate attitudes tell the same story from a different angle. The partisan gap in 2026 is enormous, with 75% of Democrats and 21% of Republicans saying warming is mostly human-caused. Pew's own description of the trend is that the Republican share has not changed. Conservative Republicans were at 15% in 2016 and 14% in 2026.^[20][21] Ten years produced no movement. Gallup's longer series shows the gap widening substantially because Democrats moved up rather than because Republicans moved down.

There is a third instance of the same pattern. The share of Americans saying it matters that a president hold strong religious beliefs fell from 72% in 2008 to 62% in 2016, and Pew noted at the time that the Republican decline predated Trump's nomination and was mirrored among Democrats.^[1]

Attitudes the argument presents as symptoms of Trumpism were in place before Trump and did not shift under him. The claim rates Misleading, not False, because one thing it points at is real. Roughly six in ten Republicans have told pollsters since 2020 that the last election was stolen, and that figure held steady across Monmouth and PRRI through 2024.^[22][23] A 2026 reading puts it at 50% of Republicans, and splits it 66% among those identifying with MAGA against 32% of other Republicans.^[24] That is a faction, not a party, and it is a specific false belief rather than a general collapse in regard for evidence.

Is the pattern one-sided?

The argument treats factual denial as a property of one coalition. Measured directly, it is not.

In a Gallup probability-panel survey in August 2021, 41% of Democrats estimated that at least half of unvaccinated people who caught COVID had been hospitalised, against 22% of Republicans. The true figure was under 1%. Only 2% of Democrats gave a correct answer, against 16% of Republicans.^[25] The same instrument reversed on vaccine efficacy, where Democrats were substantially more accurate. The error tracked what the answer would mean politically rather than which party was answering.

Consumer sentiment makes the point without any ambiguity. On the University of Michigan index, Democrats sat at 91.6 and Republicans at 50.2 in October 2024. By December 2025 Democrats were at 34.6 and Republicans at 90.6, describing the same economy.^[34]

That said, an asymmetry does exist and it would be dishonest to stop here. Linked Facebook data found 18.1% of Republicans shared at least one fake news article against 3.5% of Democrats, though more than 90% of all users shared none and the age effect was stronger than the party effect.^[28] A 2024 study in Nature found accounts using Trump hashtags shared markedly lower-quality news even when quality was rated by politically balanced panels of ordinary people, which defeats the usual objection that fact-checkers are partisan.^[29]

The defensible synthesis is that motivated misperception is general and reverses with the political valence of the question, while the supply of viral political falsehood and measured susceptibility to it have been asymmetric. Both halves are needed. The claim as stated, that this is a distinctively Republican trait, rates False.

One place the argument is right and its critics are wrong

The standard rebuttal to any of this is that partisans do not mean what they tell pollsters, that survey answers are cheerleading. On this particular belief that rebuttal fails, and the failure is well established.

A 2024 study across five surveys and 18,833 respondents used six designs, including bets with real money, to test whether stated belief in 2020 election fraud was expressive. It found none. The largest estimate was 2.3 percentage points and three of the six took the wrong sign. Around 80% of Republicans in November 2020, and more than 90% by July 2021, were revealed as sincere.^[27] A 2026 meta-reanalysis of raw data from 44 studies covering 242 questions found that accuracy incentives shrink measured partisan bias by about a quarter, and concluded that expressive responding is a modest component of partisan bias, which is itself a modest component of belief.^[26]

People who say they believe the election was stolen mostly believe it. The dismissal rates False.

The strongest claim, and the weakest evidence

The argument closes by saying that many voters were willing to act against their own self-interest to harm their enemies, driven less by gain than by the satisfaction of seeing opponents suffer. That is the most striking sentence in it and the least supported.

The groundwork is solid. Warmth toward one's own party has hovered in the 70 to 75 range for decades, while ratings of the opposing party fell from about 48 degrees in the 1970s to roughly 20 today.^[33] Displeasure at a child marrying across party lines rose from around 5% of Republicans and 4% of Democrats in 1960 to 27% and 20% by 2008.^[36] The two ends of that comparison use different answer scales, which is a known weakness in the most-quoted statistic in the field, and the direction still holds. Something large has happened to how partisans feel about each other.

Feeling is not action, and the behavioural claim has a gap in the middle of it. The experiments that find partisan harm measure one of two things. Either the harm pays the person doing it, as in allocation games where giving less means keeping more, or it costs them nothing at all.

The cleanest of the zero-cost designs is worth describing, because it is the strongest evidence the argument has. In a 2026 study of 1,842 American partisans, one player could hand a second player money. The second player then decided, at no cost to themselves either way, whether the first player also got paid. Against a co-partisan, 8.1% chose to leave the other person with nothing. Against a political opponent, 34.0% did.^[37] Those who did it also rated the choice as ethically appropriate. Across all three studies, out-group discrimination appeared in eight of eight measures while in-group favouritism appeared in only three of eight. Partisans were not rewarding their own so much as punishing the other side.

An earlier and much-cited pair of games found the same asymmetry. Measured against a control in which the other player had no stated party, the penalty applied to an opposing partisan ran to roughly twice the bonus given to a co-partisan, and describing the other player as an Independent produced no effect at all.^[31] One caution applies to that design and not to the 2026 one. In an allocation game, handing an opponent less means keeping more, so the penalty is money the participant pockets. Self-benefiting discrimination and self-sacrificing spite look similar in a results table and are different claims.

That is gratuitous spite, and it is real. It is still not the claim. Nobody in that experiment gave up anything.

The stronger claim has been tested directly, twice. Lelkes and Westwood built a study series around the top of the prejudice hierarchy, including endorsing tear gas against an opposing group and selectively investigating an opposing official, and found that the most affectively polarised partisans were no more likely to intentionally harm the opposition than the least.^[30] The same team re-ran it through the 2016 campaign and its aftermath and reported no general increase, concluding that the limits identified earlier remained in place.^[38]

A 2024 experiment put self-interest and partisanship into direct competition and self-interest won outright. When the consequences for a partner were hidden, giving respondents an excuse, selfish choices rose to 49.5%, and the rate of taking that excuse was identical whether the partner was a co-partisan or an opponent. Group favouritism shrank rather than grew.^[39]

The related dispute about voting against material interest points the same way. Larry Bartels, using the definition Thomas Frank himself proposed, found that low-income white voters became more Democratic across the period in question, by 4.5 points nationally and 11.2 points outside the South, and that the decline among non-college white voters was overwhelmingly Southern, at 19.7 points against 1.0 point elsewhere.^[32] Frank's own mechanism was delusion rather than malice, so citing that literature for a spite thesis misreads both sides of it.

The supportable statement is that partisans will impose costs on opponents when it brings them no benefit. The stated claim, that they accept costs to themselves for the pleasure of it, rates Unproven. What would settle it is a well-powered study in which partisans forgo real money, rather than merely decline to transfer it, to reduce an opponent's payoff. None was located, which is a limit of the search rather than proof that none exists.

The proposed mechanism fares worse than the claim. When voters were told a candidate favoured policies that would disproportionately harm the other party, stated willingness to vote for that candidate fell by 21.2 points against a no-information control, and by 17.6 points against a candidate promising to help both parties. The authors' summary is that cruel candidates are disliked on average, and that the promise of cruelty appeals only at the very highest levels of measured schadenfreude.^[40] Separate survey work puts partisan schadenfreude and support for partisan violence at 5% to 15% of partisans, against 40% to 60% for milder forms of moral disengagement.^[41] Appetite for opponents' suffering sits in a tail of the electorate, not at its centre.

How solid is the underlying measure?

The argument leans on the idea that partisans now loathe each other, so the durability of that measure matters. A substantial body of work says the standard instrument overstates it.

When researchers randomised what respondents were actually rating, out-party voters scored about 29 while party elites scored about 25 and the party in the abstract about 23, with elites and the party statistically indistinguishable. More than 80% said they were at least somewhat comfortable with an out-party friend or neighbour.^[42] People picture politicians when asked about the other party. Related work found respondents believe 69% of out-partisans are ideologically extreme when the true figure is 38%, and rate a moderate out-partisan who rarely discusses politics at 56 degrees, which is warmer than neutral.

A second line finds much of the measured hostility is dislike of partisanship itself rather than of the other side. An in-law who rarely discusses politics reduces unhappiness by about 5 points, and only around 35% would be happy about a co-partisan in-law.^[43]

Most damaging to any claim that this animosity drives behaviour, a 2023 study moved affective polarization by 14.3 degrees, roughly three decades of secular change, then measured downstream effects on democratic norms, electoral accountability, adopting party positions, and willingness to compromise. Of 75 test statistics across those areas, three reached significance, about what chance would produce.^[44] And survey estimates of support for political violence appear inflated roughly sixfold by inattentive respondents, with a corrected median around 2.9% against a published median of 18.5%.^[45]

These critics are not saying the phenomenon is invented. Each of them says plainly that affective polarization is real and has increased. What they show is that the gap between disliking the other party and acting against it is wider than the argument assumes.

What a fact-check cannot settle

Several of the argument's load-bearing sentences are not rated above, and the reason is worth stating plainly. When it says supporters hated the idea that science contradicted their religion, or that denial was a strategic decision rather than ordinary motivated reasoning, or that they were driven by the satisfaction of seeing enemies suffer, it is describing interior states of millions of people.

No poll reaches motive. Surveys reach stated belief, experiments reach behaviour under specific conditions, and both leave the question of why substantially open. A verdict badge on that kind of sentence would be false precision. Those clauses can be argued but they cannot be checked, and a fact-check that pretended otherwise would be doing something other than fact-checking.

The bottom line

An argument can be built almost entirely from true statements and still assert a relationship between them that the evidence does not support. That is what happened here. The alignment it describes is real and overwhelming. The attitudes it treats as consequences of that alignment either predate it, moved everywhere at once, or moved after it. PRRI stopped asking its morality question in September 2020, and a check of every American Values Survey topline since confirms it has not returned, so the central statistic in this debate has not been measured in six years.^[35]