A new NBC/Wall Street Journal poll shows that Americans favor Democratic control of Congress by a margin of 50%-38%. Ten other polls conducted in the last ten days have found similar results, with the Democrats’ advantage ranging from 9%-18%. The last time the Democrats had this kind of lead was in the 2006 and 2008 elections. The Democrats picked up 52 seats in those two elections.
What should be even more chilling to Republicans, as it relates to their long-term prospects, is that millennials preferred the Democratic Party by a margin of 68%-17%. I have never seen a 51% margin in a party preference poll in any age bracket before. Millennials are now the largest voting bloc in the country, but, like previous generations, have been slow to start voting. But that may be changing. The Alabama Senate election results indicate that some of the Republicans’ positions on issues like gay marriage and climate change are beginning to motivate millennials.
Republicans immediately attempted to discredit the polling, citing the 2016 Presidential election poll. It has become a common retort to any polling unfavorable to Republicans to claim that polls cannot be trusted because they badly missed the 2016 results. While there is no question that polling is as much an art as a science, it is foolhardy to completely disregard polls that are based on representative samples (as opposed to self-selected internet polls or “push” polls).
The trick in any poll is to try to match up the results from the interviews to a prediction about who will show up to vote by adjusting the actual results to more closely reflect the anticipated turnout. Frequently the models used for this adjustment are based on historical turnout. And that is where most polls got it wrong on the 2016 election.
Over the last 20 years or so, whites without a college education have gradually become increasingly alienated from both political parties and as a result their participation has declined. However, in Trump they found someone who articulated many of their frustrations. They were energized and turned out in larger numbers than most pollsters predicted. This is precisely the type of basic change in voting behavior that is the most problematic for pollsters.
However, not all pollsters got it wrong. One widely reported outlier was the USC/LA Times poll that consistently showed Trump doing better. For a more detailed described of how their methodology differed, [click here].
So, how far off were the 2016 polls? Turns out, not that much. Real Clear Politics is a website that aggregates poll results. It listed 61 polls in the month before the election that included all four candidates. Here is how the average of those polls stacked up against the actual results. [Click here to review a list of the polls.]
So, where the pollsters got it wrong was underestimating Trump’s support. The magnitude of the miss was relatively small, less than 3% outside the margin of error. But because the election was so close, this relatively small miss was critical to the outcome. (Note: It is important to remember that a 40,000 vote swing in Michigan, Wisconsin and Pennsylvania would have changed the outcome of the election.)
The 2016 polling results simply do not support the current narrative among Republicans that polling can be disregarded or that it dramatically understates their support. When the polling shows a 12% spread (or a 51% spread!), trust me, the Republicans have a problem. Of course, polling is a snapshot in time and it does not necessarily predict the future. The attitude of the American people may well change by next November. But if the election were held today, the Republican Party would be in a world of hurt.
I make this observation with no sense of celebration, but rather with great anxiety. For as disgusted as I am with the child-molester-tolerant, Russian-infatuated, deficit-addicted party the GOP as turned into, the Democrats, who cannot do basic math, are even worse. Hopefully we will soon have alternatives to these to out-of-touch, fringe-pandering, corrupt parties. But I fear that may not be as soon as 2018.