Many Republican senators, watching the harrowing footage of the Jan. 6 Capitol insurrection played at Donald Trump’s second impeachment trial, were moved to tears, presumably remembering their own experiences that day. Yet it did not persuade many of them to vote to convict the former president on the charge of inciting the insurrection. Even though many prominent Republicans, including Sen. Mitch McConnell (Ky.), appeared to want a clean break with the president after the insurrection, those who turned on Trump have been rebuked by the party’s grass roots. And Trump has threatened to recruit and support primary challengers against Republicans who do not line up behind him.

So were the votes against conviction motivated by a desire to win primaries and, therefore, reelection? Our research shows that the situation is more complicated than that. The Republican rank and file is deeply divided over Trump and his false claims about a stolen 2020 election. This creates a minefield for Republican members of Congress.

Here’s how we did our research

With the support of the University Center for Human Values and the Data Driven Social Science Initiative at Princeton University, we launched the Election Legitimacy Tracking Survey seven days before Election Day, an online survey conducted through Qualtrics that measured perceptions of the election outcome. Between Oct. 27 and Nov. 20, we collected 500 nationally representative responses per day. After Nov. 20, we switched to collecting 500 responses every Tuesday and Friday. We paused around the holidays and restarted the survey on Jan. 8, wrapping things up on Jan. 29. All told, we have 40 days’ worth of data from 20,000 respondents.

Like other pollsters, we find that just over half of Republicans do not believe that President Biden really won the election. This refusal to accept Biden as the legitimate winner remains even through such newsworthy events as the Jan. 6 insurrection and Biden’s inauguration.

To better understand the political implications of this, we included a survey experiment in the polls conducted Jan. 12 to 29. We asked respondents who identified as Republican to imagine that they were choosing between two Republican candidates. We then showed a table like the one below and randomly varied attributes like profession and gender. More important, we varied whether the candidate took a position on the outcome of the 2020 election: Either they believed that Biden won and would have certified the result for him, or they believed that Trump won and would have not voted to certify the election for Biden. We asked respondents to look at three pairs of candidates so that we could ensure that all of the possible combinations would have been seen by some of our respondents.

Of course, people don’t really select candidates from a table of attributes. However, this design allows us to approximate the effect of a candidate publicly accepting Trump’s false assertion that he won the 2020 election.

Let’s begin with the overall effect of a Republican candidate rejecting Trump’s claims about the election. Even after accounting for profession, religion, gender and so on, a candidate who asserts that Trump really won the 2020 election did better than those who say that Trump lost. On average, that candidate gains about six percentage points against the opponent.

But this is what happens on average. What happens when we break out the results based on whether the respondent believes that the 2020 election outcome was legitimate? As you would expect, Republicans who thought the election was stolen from Trump reward candidates who say he won, while those who thought Biden won punish candidates who say Trump won. Perhaps more important, we see an asymmetry.

The Trump-really-won Republicans reward Republican candidates who agree with them more than Biden-won Republicans punish the candidates who disagree. To put it differently, Republican voters who continue to believe that Biden stole the election appear ready to punish Republican candidates who say otherwise. Republican voters who accept the Biden victory give less weight to the issue as a whole.

That’s the minefield facing Republican members of Congress. If Republican primary electorates in states and congressional districts look like our national sample, embracing the truth about the 2020 election could be political suicide.

Of course, in districts dominated by Republicans who reject the myth that Trump won, candidates who accept Trump’s loss could be rewarded. In other words, the 2022 elections could deliver a Republican congressional caucus that is more deeply divided than it is already.

Whatever happens, our data suggest that the Trump-won faction has the upper hand. Republican politicians and officials appear to understand that they take on the former president at their political peril.

Kevin Arceneaux (@VinArceneaux) is the Thomas J. Feeney Jr. Professor of Political Science at Temple University.

Rory Truex (@rorytruex) is an assistant professor of politics at Princeton University.