President Trump’s election came as a shock to many because 2016 polls in Pennsylvania and several critical battleground states in the Midwest consistently underestimated his level of support, missing late-deciding voters and failing to account for huge differences in opinion based on education levels.
That was the conclusion of a review by the American Association for Public Opinion Research Thursday. It found that national public polling was “generally correct and accurate by historical standards” in reflecting the popular vote, but some state-level polls were off the mark, notably surveys in Pennsylvania, Michigan and Wisconsin, the states that cinched an electoral college victory for Trump.
Given the perception that Democrat Hillary Clinton was heavily favored, the polling industry itself came in for a battering in the court of public opinion, and the professional association commissioned the what-went-wrong study.
In Pennsylvania, surveys underestimated Trump’s support by 4 to 5 percentage points. The polls in Michigan were 3.5 percentage points off in assessing Trump’s eventual vote. In Wisconsin — which gave its electoral votes to a Republican for the first time since 1984 — the polls low-balled Trump by 6.5 points.
He carried the three states by a combined 77,740 votes, or about one-half of a percentage point.
The report also throws cold water on Clinton’s contention that she would have been president if it hadn’t been for the surprise announcement Oct. 27 from FBI Director James Comey that agents had found previously uknown emails from her private server.
"The Comey letter had an immediate, negative impact for Clinton on the order of 2 percentage points," it says. "The apparent impact did not last, as support for Clinton tended to tick up in the days just prior to the election."
National polls were pretty close to reality in the end, the report says. Clinton was leading by an average of 3 percentage points heading into Election Day in these surveys — and she ended up winning the popular tally by 2.1 percentage points, or about 3 million votes.
“While the general public reaction was that ‘the polls failed,’ we found the reality to be more complex,” the report concludes. “Some polls, indeed, had large, problematic errors, but many polls did not.”
Exit polls found that voters who decided in the last week of campaign broke heavily toward Trump. The report calls that “good news for the polling industry” because it would mean many state-level surveys were accurate when they were taken and not inherently flawed.
It also says that a number of state polls did not correct for over-sampling of college-educated voters, giving an advantage to Clinton. Voters without college degrees were much more likely to support Trump.
Here is a Q and A from Pew Research about the findings and implications of the report.