Trump's Big If: Why Polls May Underestimate His Support
Polls may be underestimating Donald Trump's support, according to intriguing new research that says the Republican front-runner benefits from a "social desirability bias"—some people who plan to vote for him are too embarrassed to admit it.
The study, published Monday by Morning Consult, found that Trump fares about six percentage points better among likely Republican voters in online polls than when a pollster is speaking by phone to a live human being. Moreover, the report indicated, the higher an individual's educational attainment, the greater the likelihood that the respondent wouldn't admit on the telephone to supporting Trump.
"Much work remains to better understand which types of polls are actually right in predicting Trump's support levels, but a key implication of the study is that many national polls may be underestimating Trump's support levels," the study concludes.
The study raises "social desirability bias" as a reason for this discrepancy. Under this theory, private online polls are more indicative of what likely voters will do in the privacy of the voting booth. There have been instances of this in past elections, most famously the "Bradley effect."
This was named after Los Angeles Mayor Tom Bradley, an African-American who led the polls going into the 1982 election for governor of California, but lost. Analysts say the polls were skewed due to voters not wanting to admit they opposed an African-American candidate. A "Trump effect," if true, would find pre-election polls influenced by supporters too shy to publicly attest to supporting a blustery white candidate whose critics have called him a racist.
Earlier polls have shown Trump to be particularly out of favor in well-educated communities. But the study suggests many of these well-educated adults privately support him, even if they won't admit it. "Among adults with a bachelor's degree or postgraduate degree," it found, "Trump performs about 10 percentage points better online than via live telephone."
A new Quinnipiac poll released Tuesday, featuring live phone interviews, found that 50 percent of American voters say they would be embarrassed to have Trump as president. The Morning Consult study raises a question: How many would say that if they weren't talking to a live person? They survey also found Trump leading the field nationally with 28 percent, a four percent cushion over second-place Ted Cruz.
There are reasons to be skeptical that Trump's support is being underestimated. First, online surveys are generally viewed by many pollsters as less reliable. Second, polls suggest Trump's supporters are historically less likely to vote, leaving considerable uncertainty as to whether they'll show up to caucus or vote in a primary. The best predictor of future voting, many pollsters say, is past voting. Indeed, many believe Trump's support is being overestimated—that many of his supporters will jump ship, or stay home.
Ken Goldstein, a political scientist at the University of San Francisco who works as a polling analyst for Bloomberg Politics, analyzed the reasons for Trump performing better in online polls than in interviews.
"Is that because of the composition of the sample? Or is that because people are embarrassed to say they're for Trump? And the Morning Consult study pretty unambiguously showed that it was [about] social desirability," Goldstein said on Bloomberg TV's With All Due Respect Monday. "It probably means [his support is] understated when you go into the sanctity of a secret ballot."
The Morning Consult study identified 2,937 Republican or Republican-leaning voters and randomly assigned them to one of three methods: live telephone, traditional interactive telephone and online. Trump had 38 percent support among online respondents, 36 percent of those answering an automated call and 32 percent when queried personally by a human pollster.
To continue reading this article you must be a Bloomberg Professional Service Subscriber.
If you believe that you may have received this message in error please let us know.