CNN is skewing polling data to show the majority of Americans disapprove of the way President Donald Trump is handling the presidency.
The network published a poll Sunday that claims 35 percent of American adults approve of how Trump is performing as president. Those results are in stark contrast to a Monday Rasmussen Reports poll that found 49 percent of respondents approved of how the president is managing the White House.
What accounts for a roughly 15 percentage point disparity?
The Daily Caller News Foundation obtained a full copy of CNN’s poll, which was conducted in conjunction with SSRS, a survey and market research firm. The data include a breakdown of respondents by age, party affiliation, income and other demographic factors. It also includes the methodologies used in conducting the poll, although some information is missing, like how CNN/SSRS chose to sample respondents.
After a review of the CNN data and methodologies, TheDCNF found a number of factors that could account for the 15 percentage point disparity between CNN’s and Rasmussen’s results.
Oversampling Democrats And Independents
CNN polled a total of 1,016 adults from Feb. 20-23 nationwide and had respondents self-identify their political affiliations. Thirty-three percent of respondents identified as Democrats, 44 percent as Independent or members of another party and 23 percent as Republicans.
The spread of political ideology could be a problem here, because it might not be an accurate representation of how U.S. adults would describe themselves politically. Looking at the most recent Gallup polling on how Americans self-identify politically, CNN’s figures are off as much as 14 percentage points for Republican representation.
Forty-four percent of U.S. adults self-identify as Democrat or independents who lean Democrat, according to a December 2017 Gallup poll. Roughly 37 percent of Gallup respondents self-identified as Republicans.
Polling All Adults Versus ‘Likely Voters’
CNN polled the opinions of “adults,” ages 18 and up, however, Rasmussen polled “likely voters.” Choosing one method or the other can lead to different results.
Polling “adults” includes a significant number of respondents in a given sample that either rarely or never vote. Adults that are not politically active, or rarely politically active, are less likely to follow public policy debates as closely as those who are committed voters.
Using “likely voters,” some argue, like Kyle Kondik, managing editor of “Sabato’s Crystal Ball,” is a better barometer for measuring presidential approval ratings.
“If I were to start doing a poll of presidential approval, I personally would do registered voters,” Kondik told TheDCNF in July 2017. “If you’re looking specifically at trying to figure out the electoral effects, you’re probably better off doing registered voters or likely voters.”
Historically, national polling houses use “adults” rather than “likely voters” when polling presidential approval.
For example, Gallup polled Trump’s approval rating using “adults” instead of “likely voters” from January 2017 to July 2017. Gallup’s approval rating was consistently below Rasmussen’s results, sometimes as much as 11 percentage points.
Reuters polled both “adults” and “likely voters” on their approval of the president since Trump took office in January 2017 and the results exhibit the same pattern. Respondents who voted in the 2016 presidential election cycle had more favorable views of the president than respondents who chose not to vote, the news agency found.
Rasmussen, unlike CNN, also does polling on a rolling basis, a factor the polling house argues makes their results even more accurate. CNN polls Trump’s approval rating every month, collecting data from respondents over a three-day period and reports the findings.
“Our poll is ongoing. The presidential tracking number that you see every morning at 9:30, it is a rolling survey of three nights,” Francis Coombs, managing editor at Rasmussen, told TheDCNF. “That is 500 likely U.S. voters, every night. You tend to see continuity, or slow, incremental ups and downs in the results.”
“These snapshots, like CNN does, where they drop in and survey for a specific period. You don’t see any context or continuity. It really doesn’t give you a measure of how the president is doing, regardless of who is in office,” Coombs told TheDCNF. “Polling American adults means just anybody. Anybody who answers the phone is included in CNN’s poll.”
Polling presidential approval in a snapshot manner can also skew results, like following a national tragedy or big news event.
Other Potential Errors
CNN did not provide TheDCNF enough detail about how they sampled respondents to provide an accurate critique of their methods, but there is some information that could call the poll’s results into question.
CNN and SSRS did not, independent or otherwise, respond to multiple requests for comment regarding how they chose to pick their random samples.
Here is everything TheDCNF was provided in terms of methodology:
A total of 1,016 adults were interviewed by telephone nationwide by live interviewers calling both landline and cell phones. Interviews were conducted in English and Spanish. Interviews were conducted February 20-23, 2018. Among the entire sample, 33% described themselves as Democrats, 23% described themselves as Republicans, and 44% described themselves as independents or members of another party.
All respondents were asked questions concerning basic demographics, and the entire sample was weighted to reflect national Census figures for gender, race, age, education, region of country, and telephone usage.
Results for the full sample have a margin of sampling error of +/-3.7 percentage points. For the sample of 909 registered voters, it is +/-3.9 percentage points.
Crosstabs on the following pages only include results for subgroups with enough unweighted cases to produce a sampling error of +/-8.5 percentage points or less once adjusted for design effect. Some subgroups represent too small a share of the national population to produce crosstabs with an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a design-effect adjusted sampling error larger than +/-8.5 percentage points are not displayed and instead are denoted with “N/A.”
The CNN poll has a margin of error of plus or minus 3.7 percentage points, which is nearly one percent higher than a typical margin of error for a poll with a sample size of 1,000 and a 95 percent confidence interval. Typically, a sample of a similar size with the same confidence interval will have a margin of error of plus or minus 2.5 percentage points.
Other potential errors include the use of “live interviewers.” If an interviewer’s intonation or inflection was different, some of the results could be called into question. Similar errors could stem from a question being worded differently in an interview.
Another point of concern with having a live interviewer, as opposed to a recorded, robotic interview like Rasmussen uses, is that respondents might feel the need to curb their responses or lie. FiveThirtyEight has pointed out a number of reasons using live interviewers can cause varying results between polls.
CNN and SSRS did not respond to requests for comment as to how they guarded against interviewer biases.