Reuters published a story on Thursday about a recent Reuters/Ipsos poll with this blazing headline: “One in eight people who voted for Trump having second thoughts.”
“About one in eight people who voted for President Donald Trump said they are not sure they would do so again after witnessing Trump’s tumultuous first six months in office, according to a Reuters/Ipsos poll of 2016 voters,” Reuters began its report, adding its own left wing bias of interpretation [emphasis added].
A detailed analysis of this latest Reuters/Ipsos poll, however, shows that the data within the poll does not really support either the headline of the story or the misleading first sentence.
The headline says one in eight Trump voters are “having second thoughts,” which implies they would now vote against Trump.
The more accurate headline, based on the actual Reuters/Ipsos Poll results would be this: “One in twenty people who voted for Trump would not vote for him again, a significant improvement from one in eight back in May.” (One in twenty is five percent. One in eight is 12 percent.)
The associated detailed data analysis provided by the Reuters/Ipsos poll proclaims that “cracks emerge in Trump’s base.”
But the actual data within the poll shows that Trump’s standing with his base actually improved during the past two months.
The Reuters’ story is based on the answers to one question taken in the Reuters/Ipsos poll conducted of 2016 Trump voters in July. Reuters/Ipsos asked the same question of 2016 Trump voters in a poll in May, and Trump’s standing among those voters during the two months period improved by six percent.
In May, 82 percent of Trump voters said they would vote for him again. In July, that number jumped to 88 percent.
In the July poll, of the remaining 12 percent of 2016 Trump voters, only five percent, however, said they would either vote for others over Trump (Hillary Clinton, Gary Johnson, or Jill Stein), or would not vote at all. Seven percent said they did not know.
Here’s a detailed breakdown of the poll question and the responses from the both the May and July polls:
The poll surveyed voters who already had told Reuturs/Ipsos on Election Day how they had cast their ballots. After several months passed, the poll asked: “If the 2016 presidential election were held today, who would you vote for?”
In May, here’s what Trump voters said:
82 percent Would vote for Trump again
9 percent Would not vote
5 percent Don’t know
4 percent Would vote for others (Hillary Clinton, Gary Johnson, or Jill Stein)
In July, here’s what Trump voters said:
88 percent would vote for Trump again
1 percent would not vote
7 percent don’t know
4 percent would vote for others (Hillary Clinton, Gary Johnson, or Jill Stein)
Reuters asked that same question in the July poll about Hillary Clinton:
In July, here’s what Clinton voters said:
86 percent would vote for Clinton again
3 percent would vote for Trump
8 percent don’t know
3 percent would vote for others or not vote
Reuters did not report whether they asked the Hillary Clinton “vote for again” question in the May poll, and, if they did, what the results were. It is noteworthy that the Reuters/Ipsos poll changed the possible responses in the “vote again” question depending on whether they were asking about Donald Trump or Hillary Clinton.
Reuters is the same organization that predicted on November 7, 2016, “With hours to go before Americans vote, Democrat Hillary Clinton has about a 90 percent chance of defeating Republican Donald Trump in the race for the White House, according to the final Reuters/Ipsos States of the Nation project.”
“[Clinton] was leading Trump by about 45 percent to 42 percent in the popular vote and was on track to win 303 votes in the Electoral College to Trump’s 235, clearing the 270 needed for victory, the survey found,” Reuters’ Maurice Tamman wrote and tweeted on election eve:
Reuters/Ipsos States of the Nation: Clinton odds-on winner. Trump path is white, black & Hispanic turnout shifts https://t.co/lQIvY8KQLW
— Maurice Tamman (@motamman) November 7, 2016
As Breitbart News reported in July 2016, this is not the first time Reuters/Ipsos has been called out for releasing biased polls.
“The Reuters/Ipsos polling team announced Friday that they are dropping the ‘Neither’ option from their presidential preference polls after their tracking polls showed a 17-point swing in favor of the Republican nominee Donald J. Trump, exposing the ‘Secret Trump Voters’ Democrats fear,” Breitbart’s Neil McCabe wrote:
“In a presidential campaign notable for its negativity, the option of ‘Neither’ candidate appears to be an appealing alternative, at least to participants in the Reuters/Ipsos opinion poll,” wrote Maurice Tamman, the leader of the Reuters news service’s New York City-based data mining and investigative reporting team. “Many voters on both sides have been ambivalent in their support for Democratic nominee Hillary Clinton and Republican nominee Donald Trump, complicating the task of the pollsters trying to track the race.”
Breitbart News noted the 17-point swing for Trump, which seems to have set off alarm bells at Reuters.
Tamman wrote that in the “Year of Neither,” the swing came from an underreporting of Trump support before the Republican National Convention held in Cleveland July 18 to July 21.
In September 2016, Tyler Durden wrote at ZeroHedge, “Reuters has taken some heat in recent months for ‘tweaking’ their polling methodology seemingly every time the data reveals ‘inconvenient’ results for Hillary (see our previous posts on the topic here and here). But the latest Reuters/Ipsos polling ‘tweak’ is truly amazing”:
Having run out of options for slyly “tweaking” questions and categories to sway respondents in their preferred direction, Reuters has apparnetly [sic] resorted to blatant poll tampering by altering their polling samples to include a disproportionate number of democrats.
In their latest poll, released just two days ago, Reuters found Hillary to have a 6 point lead in a head-to-head contest with Trump. But, when you dig a little deeper you find that Reuters’ polling sample included 44% democrats and only 33% republicans. Which would be fine, of course, if it had any basis in reality. But, as The Pew Research Center points out very clearly (see table below), registered democrats represent about 33% of the electorate while republicans are 29%…a modest 4 point gap versus the 11 point advantage in the Reuters sample.
Reuters provided the following description of the methodology of the July 2017 Reuters/Ipsos poll:
The Reuters/Ipsos poll was conducted online in English throughout the United States and has a credibility interval, a measure of accuracy, of about 5 percent points.
The July 11-12 poll gathered its sample from 1,296 people, including 541 Trump voters, while the May 10-15 poll gathered its sample from 1,206 people, including 543 Trump supporters. In both cases, Ipsos weighted their response according to voter profiles gathered from the U.S. Census’ voting and registration supplement to the Current Population Survey.
The Reuters/Ipsos poll is one of the few national polls that is conducted entirely online and does not include either live or automated calls to landline phones or cell phones.
One unusual aspect of such polls is that, according to Reuters/Ipsos, “statistical margins of error are not applicable to online polls.”