Putting Trump's Upset Potential in Perspective

This weekend early voting started in a number of key states, including Virginia, Minnesota, Michigan, Texas, & Georgia, all states that are going to decide the election in November for the White House, Senate, and House.  With us getting close enough that physical, tangible votes are starting to be cast, pundits are going to start to actualize numbers, and start running scenarios based on past elections or the current polls, trying to interpret, or occasionally misinterpret them.

I don't put this in the same category as the media, which always wants a horserace, and will do whatever it takes to make sure that they get one, even if the evidence is not there.  The polls tell a pretty clear, consistent picture-that Joe Biden is a sturdy favorite, a sturdier favorite than Hillary Clinton ever was, to win at least 270 electoral college votes in November.  He is not a sure-thing; Trump still has a path, though it is admittedly narrowing at this point with polling in Minnesota, Michigan, Wisconsin, and Arizona (four key swing states) all going to Biden by a heavy enough percentage it's questionable how Trump would be able to turn around one of the states & win it, and without these states (unless Trump can find a random victory in Nevada or New Hampshire), Biden is at 269 votes, with leads maintained in a number of other states & districts, and all he'd need would be to win one of them.  A 269 tie is sexy for political nerds, but it's unlikely that Biden can so handily get to 269 and not gain enough to get past it.

Most pundit coverage, though, is going to focus on this possibility that Trump might win a narrow victory or get a tie, in which case the Constitution's bizarre value of land over voters would give him the victory (why, exactly, the Constitution gives the winner of a tie to the House rather than the popular vote is a question I'd like to ask James Madison if he ever decides to start haunting me).  This makes sense, practically-there's obviously a bigger difference between the polls being right and the polls being wrong (but favoring Trump) than looking at the other direction-the polls being wrong (but favoring Biden)...Biden winning by a little or a lot isn't ultimately that big of deal for Biden.  But it is worth considering this question at least once before the election, so my ask today is-what if the polls are under-valuing Biden's support, instead of Trump's?

Obviously in this situation, Biden's going to win.  Biden wins based on the polls today, Biden could still very well win if the polls are under-valuing Trump (Trump's best case scenario is a very slim win of somewhere between 269-280 electoral college votes).  But the margin of his victory would change.  We're going to use both 538 and The Economist for our elections forecast here.  Both sites, while they differ in odds & the order of swing states' likelihood of flipping, pick Pennsylvania as their "tipping point state" (the state that will be, based on his margin of victory, the one that "won Biden the election").  538 has Biden's chances of winning Pennsylvania at 78%, while The Economist has it at 81% (I wrote this on Friday, so if they've changed over the weekend that might explain a small discrepancy, though the numbers have remained pretty static).  So essentially in order to win the White House, they're saying that Trump would need to win a state that they're giving heavy odds toward Biden.

But to prove why this is a stretch, let's take a look at Biden's states that he'd win if he were to win every state that Trump is favored in by less than or equal to what Biden is favored by in Pennsylvania.  538 has Biden favored in all states Hillary Clinton won, plus Michigan, Wisconsin, Pennsylvania (and then after the tipping point), Arizona, Nebraska-2, Florida, and North Carolina.  That's 334 electoral votes, a solid victory, and The Economist matches that percentage (though they don't break out the Nebraska/Maine districts, so we'll just assume).  But something pretty seismic happens when you assume "if Pennsylvania at 78-81% is plausible for Trump, then anything under those numbers is plausible for Biden."

In that case, we see Biden pickup Maine-2, Ohio, Georgia, Iowa, and Texas, all of which they give better odds of Biden winning than they do Trump in Pennsylvania (in the cases of both models), with Alaska just missing the cut for 538 (they give Trump an 80% chance of victory).  That brings Biden up to 413, the biggest win for a presidential candidate since 1988, and the biggest for a presidential challenger since 1980.  That's a huge difference for Trump of nearly 150 electoral college votes, and it illustrates my point-when we talk about a Trump victory, we have to assume the polling is wrong.  And if we're saying the polling is wrong, we can't just automatically assume that Trump will benefit if it's wrong-it's possible that Biden is under-performing or will have more unlikely voters on Election Day-we don't know.  But pundits & the media need to be careful about assuming that the polling will be wrong in the same way it was in 2016.  It is extremely likely that at least one state will be an upset-that we will see a state that polls favored (by whatever margin) one candidate, and they pick the other.  But in order to have conversations about Trump actually winning (and not just doing better than expected), we need to put those conversations into perspective-the likelihood that Trump pulls off a slim victory is comparable to Biden pulling off an historic victory...and if the truth is somewhere in the middle, a modest-to-good victory for Biden is still the most probable outcome in November.