PREDICTION: 314 Electoral Votes for President Trump - analysis by Jeff Cassman
Jeff Cassman | The Daily Knight
Four years ago mainstream media polling predicted an easy Clinton win with a high degree of confidence. The glee with which the media boasts of a Biden landslide is eerily reminiscent, so why do a few contrarians predict a Trump victory?
In order to answer the question we need to review how political polling works. A good pollster does not sample random people who are willing to answer an ‘unknown’ call or respond to a survey request by email; a competent pollster surveys likely voters. After all, people who haven’t bothered to register to vote, or those who’ve registered but never vote are not representative of those who actually take the time to complete and mail in an absentee ballot or show up in person to vote.
Likely voters, defined as those who voted in primaries earlier this year, or who voted in state or local elections last year, or who have a history of voting in general elections in November, are not only a more reliable predictor of who will vote, but they’re older and more conservative.
These people are more difficult to reach, however, and less likely to engage with a pollster. In order to reach a statistically significant number of people, many pollsters interview registered voters rather than likely voters and then attempt to ‘correct’ the results by giving more weight to some respondents’ answers.
For example, if the poll shows that 60% of respondents identify as Democrats, but the pollster expects Democrats to be only 55% of those voting, the Republicans’ and Independents’ votes have to be given additional weight.
This adjustment of the data is necessary not just for partisan identification, but every other correlative characteristic, e.g., age, education level, race and religion. Even a well-meaning pollster can easily make statistically significant errors when trying to adjust the polling data to be reflective of the demographics he anticipates will actually vote.
With this brief insight behind us, we can now evaluate why some veteran pollsters believe the mainstream media is making the same mistake in 2020 as they did in 2016 and why Trump is more likely to win in November than he was four years ago.
Current polling overstates Democrat enthusiasm and turnout. In 2016, the Democrat base was very enthusiastic about the Clinton campaign. That enthusiasm does not exist for Joe Biden. Enthusiasm is highly correlated with the likelihood of voting, and even a small decrease of one or two percent can have enormous outcomes (five of seven ‘swing’ states were won by less than two percent in 2016; Trump one the other two by four points). (1)
The so-called ‘shy’ Trump voter also remains a problem for pollsters. These voters secretly support Trump but don’t want their family, friends or coworkers to know. They tell pollsters they’re for Biden, or they’re undecided, when they actually intend to vote for Trump. These voters tend to be blue collar white men and they tend to live in the Rust Belt.
Pollsters also assume that millions of people who voted for Trump in 2016 have abandoned him for Biden. From Arizona to Florida, from Pennsylvania to Michigan, pollsters are predicting an enormous swing vote, in some cases by as much as 10%, and yet my data has not found a single case of a likely voter switching (although there are cases of non-voters claiming to have supported Trump in 2016 who now support Biden).
On the other hand, a statistically significant number of “Never Trumpers”, (Republicans who had said during the 2016 primaries that they would never, under any circumstances, support Trump), have changed their minds as a result of his track record on judicial nominees.
Pollsters’ models also appear to ignore the significant advantage Republicans have in new voter registration in the battleground states. In Pennsylvania alone, which Trump won by 4 points in 2016, the Republicans have a 7 point advantage over Democrats. Polls which show Biden up by as much as 13 points in Pennsylvania suggest that nearly a quarter of all Republicans in that state now support Biden. This seems highly improbable given that Trump’s approval among self-identified Republicans remains consistent and high.
Further, Trump’s ‘job approval’ number among the population at large has been very stable over the last four years, hovering around 46%. (2) If there had been a significant shift away from Trump by voters, we would expect to see it. Further, the only incumbent Presidents who have lost reelection had approval numbers of 37 and 32 (GHWB and Carter), while Truman won reelection with a job approval of just 40%.
The record-breaking voting by absentee ballot also presents challenges for Democrats, because this method of voting is far more likely to be invalidated because a signature doesn’t match, a witness wasn’t used, the form wasn’t completed properly or simply because it fails to arrive on time. Democrats are far more likely to live in fear from COVID, and are already voting absentee in record numbers, presumably because they don’t want to vote in person. (3) If lots of Democrats vote by ballot, and just a small portion of Democrats’ absentee ballots in the swing states are rejected for any reason, it could cost them the election.
Finally, an interesting anecdote; most Americans believe Trump will win the popular vote, and this measure of expectations has a perfect track record since this method of polling began in 1996. (4) Fully 56% of Americans polled believe that Trump will get the most votes, and this is likely based on their personal experiences with family members, friends and coworkers who they know are voting for Trump. It is NOT a reflection of what they hope will happen, or how they are personally voting. Clearly some Democrats who are voting for Biden think Trump will win, and it’s likely what they are seeing and hearing around them that has led them to that conclusion (or because they themselves are voting for Trump).
Interestingly, in 2016, Clinton received nearly 3,000,000 more votes than Trump did, but that’s because she ran up large margins in big cities in heavily Democrat states, like California, Illinois and New York. Trump won because he had a better distribution of voters around the country.
It’s inconceivable that he could flip millions of voters to his side and still lose. I predict Trump wins every state he carried in 2016 plus Minnesota.
Citations:
(2) https://news.gallup.com/poll/311825/presidential-job-approval-related-reelection-historically.aspx
(3) https://www.nytimes.com/2020/10/10/us/politics/early-voting-swing-states.html
(4) https://news.gallup.com/poll/321347/trump-pre-debate-job-approval-highest-may.aspx
Comments