Business Insider falsely claimed that “recent polling showed US troops had a high disapproval rating for Trump.” The poll in question, conducted by Military Times, only showed that Military Times readers had a high disapproval rating for Trump, not U.S. military personnel generally. The pollster Business Insider cited even acknowledged its survey ahd a sampling bias that favored career personnel, and may not have accurately represented the views of non-career personnel, who comprise the majority of the armed forces.
BACKGROUND: In August 2020, Military Times conducted a poll of its readers which found a “slight but significant preference for former Vice-President Joe Biden in the upcoming November election.” The poll was conducted in cooperation with Syracuse University.
THE CLAIM: A November 11, 2020 story by Business Insider alleged that “recent polling showed US troops had a high disapproval rating for Trump and were split over which candidate they would vote for.” The poll to which Business Insider referred was the August 2020 Military Times survey.
INVESTIGATION: According to Military Times, its poll only surveyed its own readers and the average age of respondents was 39. Is the average American military officer or enlistee a 39 year old Military Times subscriber? We suspected not, so decided to check.
A 2018 U.S. Department of Defense study reported the average age of U.S. military personnel to be 28 years of age, more than a decade younger than the 39 year average surveyed by Military Times.
Sampling bias can occur when a survey’s design makes it probable that some part of a population is more likely to be selected than other parts. It is explained in more detail at this University of Texas website. One famous case of sampling bias involving polling only subscribers to a publication, or set of publications, occurred in the 1936 U.S. presidential election when The Literary Digest famously predicted Republican Alf Landon would handily defeat Democrat Franklin Roosevelt based on an extensive survey conducted of persons who subscribed to magazines. Obviously, he did not. The history of this specific episode is examined at this University of Pennsylvania Department of Mathematics website.
In the case of the Military Times survey, it polled a demographic subsection of the military population, probably composed of the officer corps and senior enlisted personnel, rather than the active duty military population at-large. In fact, Military Times even notes as much. Rosalinda Maury of Syracuse University, explained that …
… the poll, which has been administered in the same way for the past four years, is not a perfect snapshot of public opinion within the entire military force. The average age of poll respondents was nearly 39 years, and the participant pool was culled and verified from Military Times subscriber lists and databases.
Underscoring the potential for sampling bias are the strange results achieved in past Military Times surveys. A September 2016 Military Times survey claimed that that 36.5 percent of active duty personnel planned to vote for Libertarian Party candidate Gary Johnson in that year’s election. In the actual election, Johnson received less than four percent of the popular vote. While it’s possible military personnel overvoted for Johnson by a 9-to-1 margin versus the public at large, such a variance would be so incredible that it seems unlikely to have escaped the notice of anyone other than Military Times.
RATING: Military Times itself acknowledged that its survey was limited to its own readers, going on to explain that those readers were probably not representative of the military at large. Business Insider made no such distinction and attached no such caveat. It falsely attributed to the Military Times survey something the survey itself never claimed. The assertion, by Business Insider, that the Military Times survey “showed US troops had a high disapproval rating for Trump and were split over which candidate they would vote for,” is, therefore, FALSE.
Status: Business Insider did not publish a correction within 72 hours of making the false claim.