A recent Morning Consult/POLITICO poll indicated that a plurality of voters think that Donald Trump, not Hillary Clinton was helped most by voter fraud. According to the poll, 35% of respondents think voter fraud helped Donald Trump and 30% think it helped Hillary Clinton. Only 18% indicated that no voter fraud occurred (another 17% said they didn’t know or had no opinion).

That’s an interesting result and, in some ways, runs counter to Trump’s narrative of the popular vote. If voter fraud helped him more than it helped Hillary Clinton, then it can’t be responsible for his being a loser in the popular vote. He just lost.

But in a more important way, it confirms Trump’s and the Republican’s narrative about wide-spread voter fraud. Regardless which candidate they think it helped in this past election, if the general public believes voter fraud is pervasive, it becomes much easier to enact legislation to fight this non-existent problem. And the problem does not exist. As nearly every credible study has shown, incidence rates of voter fraud are just infinitesimal (see the Brennan Center’s terrific report for an excellent summary of the evidence).

But the polling itself, and the reporting of it raises some concerns for me. The Morning Consult article on the results does lead with the sentence, “President Donald Trump has insisted, without evidence [emphasis added], that the 2016 election was rife with illegal voting…” But that statement could be interpreted just that Trump himself provided no evidence, not that no evidence for voter fraud exists.

While it’s true that Trump has provided no evidence, any story on this issue, including reports of survey results such as this, should go further and affirmatively state that evidence for significant in-person voter fraud does not and could not exist, because you can’t provide evidence for something that doesn’t actually occur.

I’ll go even further and suggest that an article on suggestions of voter fraud should include reference to the much more significant issue of voter suppression. It, after all, does exist. Since the wave election of 2010 and after the 2013 Shelby County v. Holder Supreme Court case, 20 states have passed laws to restrict voting. More importantly, these laws affect Black, Hispanic, and Asian voters at significantly higher rates than White voters.

Brendan Nyhan and others have researched and discussed the difficulty of challenging false information, and how attempts to rebut such misinformation can lead to a “backfire” effect. When attempt to rebut a false statement, one can actually reinforce those falsehoods with those who are motivated to believe the incorrect facts because the falsehoods support their world view. Instead, it may more effective to to frame your rebuttal in an affirmative way, rather than directly counter the false information.

This leads to my broader concern about this poll. When one asks about issues for which clear answers exist, what is our responsibility as researchers? We can study perceptions of voter fraud and we can report on what our respondents believe, but we can’t let that stand alone. But as I’ve just stated, when reporting the results, we must include the reality of the situation (e.g., widespread voter fraud does not exist; almost no cases of in-person fraud have been uncovered). Otherwise, we risk contributing to the perpetuation of false information.

But what about within the poll itself? Do we have any duty to our respondents to correct knowingly false information? Clearly, we wouldn’t want to bias the answers, but at the end of the survey, do we connect respondents to the true information? I don’t have good answers on this, but it’s an important question to consider.