Perhaps most importantly, many pollsters now evaluate their sample to adequately represent voters without a college degree. The fact that many state pollsters did not do so in 2016 is widely considered one of the main reasons that polls underestimated Trump’s support. Voters without a four-year college degree are much less likely to respond to telephone polls, and much more likely to support Trump. According to our estimates, the weighting for education could move the typical survey up to four points in the Trump direction.
Although many state pollsters still do not weigh for education, much more than four years ago. The Monmouth survey is an example. The final 2016 Monmouth Pennsylvania survey, which showed Mrs. Clinton up to four percentage points, would have shown her a two-point lead, 47 to 45 percent, if education had weighted it, according to Patrick Murray, director. of the poll. That only covers about half the difference between the actual result and the final Monmouth poll, and it’s one reason to be more confident in the new Monmouth poll.
Weighting education is not enough to guarantee a perfect result. After all, Mrs. Clinton would still have led, albeit in a rather limited way, in the final survey of Monmouth in Pennsylvania, even weighted by education. And other high-quality, education-weighted state polls, like the Marquette Law School survey in Wisconsin, still showed Mrs. Clinton with a sizeable lead in 2016. Other factors were clearly at stake.
But many of the other sources of polling error in 2016 also seem less likely to recur.
There are far fewer undecided or minor party voters now than four years ago. These voters broke in Trump’s favor, according to exit polls and post-election polls that re-contacted pre-election respondents, which helped explain some of their strength in the bottom line. Undecided voters may get closer to Trump again, but this time there are simply fewer, and therefore fewer opportunities, for polls to go wrong for that reason.
Another component of the 2016 survey error was participation. The role of participation in voting failure is difficult to generalize because there are as many likely voter models as there are pollsters. But some pollsters probably overestimated black participation by relying too heavily on Obama-era participation models. This time, relying on the latest election could easily underestimate Democratic turnout. Overall, the out-of-power party tends to enjoy a relatively higher turnout than when in power, and that could help Democrats compared to 2016.
In the case of the Monmouth survey, it turns out that participation was a significant factor. Trump would have led Mrs. Clinton by one percentage point, 45 percent to 44 percent, in the final poll from Monmouth University, Pennsylvania if it had been weighted by education and if the sample of likely voters had consisted only of those who ultimately voted, according to voter records. (Mr. Trump won the state by 0.7 of a point.)
Similarly, the final poll by the New York Times / Siena College of North Carolina, which was conducted over the weekend before the election and found the candidates tied, would have shown Mr. Trump ahead by four percentage points (won by 3.6 points) if the Sample of likely voters had consisted only of validated voters. These kinds of findings led many pollsters to conclude that the survey investigation did not fundamentally break after the 2016 election, and led many to redouble their efforts rather than quit the company.
