Election predictions are more of an art than a science. Sure, 80% of state results are a given, but those remaining 20% are usually decisive. This year is no different. Most forecasts point to a Biden victory. Of course, most forecasts pointed to a Clinton victory four years ago, so why believe forecasts this year?
Forecasts take polls of usually a few thousand voters and then extrapolate the results. They skew their samples to certain demographics. For example, if a poll had 20% respondents of African Americans but pollsters know that African Americans tend to make up 10% of the voters on election day, they may cut in half the value of the results from their sample. This is an attempt to correct for sampling errors in the original polling sample.
Of course, to make these corrections, pollsters must guess at voter turnout based on demographics. Pollsters often guess wrong. For example, white non-college educated voters turned out in record numbers in 2016, giving Trump a surprise victory. Some observers noted the enthusiasm of Trump rallies and predicted this higher turnout. Most pollsters ignore information like that which was hard to quantify.
This year, we are seeing much higher turnout for traditional Democratic voters. Although they are not showing up at rallies, we are seeing record early voting, and voters standing in long lines to vote early. We are also seeing record returns of mail in voting thanks to many state changes that encouraged mail in voting this year. Both of those trends tend to favor Democrats. I would not be suprised if there were five to ten million more total votes this year than in 2016. A total turnout of 135 million voters, as opposed to 129 million in 2016, would almost certainly favor Biden.
Also, in 2016, much of the undecided vote when to Trump in the last two weeks of the election. Much of this is blamed on the decision of the Obama Justice Department to re-open the investigation into Clinton's emails in late October. It made the discussion all about an issue that hurt the Democratic candidate. This October, there were attempts to push a story about some emails about Biden's son Hunter which indicated corruption. That story has been far less persuasive to voters, probably because it is coming from the other campaign rather than a more objective source. Instead, most discussion has remained on the Covid-19 pandemic, which is again reaching record levels, and which impacted the White House directly with infections in October. Focus on that issue can only hurt Trump and help Biden.
Note that even if Trump manages to win five of those seven states, Biden would still win the election by a more narrow margin. If Biden wins Pennsylvania, where he has had more than a 5% margin in most polls, the election is probably over Trump would have to win the others, including Michigan and Wisconsin where Biden is an 8 point favorite. Trump must win the closer states of Arizona, Georgia, Florida, and North Carolina where Biden is up 1-4 points. He must also win Iowa and Ohio where the candidates are in a statistical tie. Even with all those wins, Trump would also have to win either Michigan or Pennsylvania to put him over the top. In short, there are few realistic scenarios for a Trump victory.
All of that said, there may not be a winner called on election night. Many key states may be too close to call, and may take days to count all the paper ballots submitted this year. Trump may show an early lead from in person voting, then see that lead disappear as paper ballots are counted. Democrats are tending to vote by paper in much higher numbers than Republicans. This, unfortunately, will play into Trumps assertions that paper ballots are fraudulent and that the election was somehow stolen from him. If a couple of states are close and determinative of the elections, expect post-election litigation and political wrangling that will make Florida 2000 look like a cotillion.
I not only predict litigation arising out of the election, I expect violence too, especially from Trump supporters if their candidate does not win.
ReplyDelete