ANALYSIS

The polling industry took a beating after the 2016 election as a stunned political and media class looked for who to blame for the shock they felt upon the election of President Trump.

In particular, the “hidden Trump voter” emerged as a key explainer for pundits tackling “what went wrong,” and continues to be something Trump supporters will point to when claiming that polls are undercounting Republicans and supporters of the president.

The fact is, the national polls were fairly accurate in 2016 and were perhaps even more accurate than in 2012.

RealClearPolitics’ final aggregation of several national polls showed Hillary Clinton ahead of Donald Trump by 3.2 percentage points. As it turned out, Clinton wound up with a 2.1 point lead in the popular vote (but lost the presidency to Trump in the Electoral College).

Still, the polling industry has spent the last 18 months trying to understand how to do its job better in order to build up trust that was lost in the post-election frenzy.

Last month in Denver, the American Association for Public Opinion Research, or AAPOR, convened its annual conference at the Sheraton Denver Downtown Hotel, and brought extensive analysis of fresh data to bear on the questions that continue to haunt the industry.

I arrived in Denver looking to sort myth from reality. Are there really “hidden Trump supporters” out there being missed in polls, and if so, how do we correct that problem?

There are many different ways to interpret the “hidden Trump supporter” phenomenon.

First, there’s the assertion that people who like Trump are just less likely to be called or to take a poll at all, the “missing Trump supporter” theory.

Second, some claim that Trump supporters will take polls but understate their support for the president, either because they intentionally mislead due to pressure to conceal their views, or because they align with the president but choose to say they are undecided, the “undercover Trump supporter.”

For the last three decades, long before Trump ran for president, polling response rates have fallen precipitously. Today, typically a tenth or fewer of those contacted for a poll agree to take it.

But even as response rates fell, polls remained mostly accurate. Pollsters had been extremely lucky that the small sliver of people who did take polls have not differed in dramatic fashion from those who didn’t, except on metrics like civic engagement. (People who are more likely to vote are more likely to take a poll, in general, and so this bias is a potentially helpful one if you are trying to just study voters.)

After the 2016 election, I worried greatly that pollsters’ luck had run out. Even if the 2016 national polls were “right” and had just been misinterpreted, the intense scrutiny of polling could easily become a self-fulfilling prophecy.

If Republicans viewed pollsters as a hostile force and if response rates among Republicans fell disproportionately, this would be an enormous challenge for the industry, as that supposed polling bias would become reality.

The good news is that this isn’t happening. Researchers with CBS News took a look at the lists of who they were calling up for their surveys. They know the political leanings and demographic characteristics of those who said yes, but crucially, they were able to match up many of the “did not answer” or “refusal” phone numbers to voter lists and databases of key demographic information.

They found that phone numbers at Republican households were just as likely to refuse a survey, or to complete a survey, as phone numbers at Democratic households.

Furthermore, researchers with Reuters/Ipsos wanted to see if the reason why the polls move is because Republicans stop taking polls during news cycles that were bad for the president.

Instead, what they found was there was not much difference in someone’s likelihood to take a follow-up survey if they were Republican, suggesting movement in the polls is more about people changing their minds rather than Republicans checking in and out of the survey-taking process.

The bottom line is that Republicans are no less likely to take surveys than Democrats, and there has not been evidence of a change in that pattern, even since the election.

This mostly debunks the myth of the “missing Trump voter” and is a much-needed dose of good news for the polling industry as the 2018 midterms approach.

There is, of course, still the “undercover Trump voter” theory which should absolutely be a focus of future research.”Undercover Trumper” is the theory that members of Trump’s own 2016 polling team will point to when asked, noting that some who agree with the president on other issues nonetheless “do not volunteer their support or leanings,” and that the key may be developing other metrics to best suss out a poll respondent’s true feelings or future behavior.

But the good news is that people on both sides of the aisle are still talking to researchers in the first place. Pollsters may not be the most popular people around, and the polls may not be well trusted, but people can’t seem to stop paying attention to them and – thankfully – Republicans haven’t stopped taking them.

Kristen Soltis Anderson is a columnist for the Washington Examiner. She is the co-founder of Echelon Insights, a D.C.-area-based polling and data agency, and regularly brings her pollster expertise to CNN, Fox News and MSNBC.