Going into Election Day, the race is as follows: The most straight-forward conclusion to be drawn from the publicly available polling data is that Joe Biden will be elected president by a solid margin. Because pollsters and pundits were wrong in 2016 and polls could be understating President Trump’s support, most prognosticators are more cautious about predicting a Biden win than they were in predicting a Clinton wipeout four years ago.
But as the final votes come in, there’s one thing I keep coming back to: All Biden has to do to win the presidency is to do a smidgen better than Clinton did.
Biden does not have to win Texas or Georgia to be president. He does not have to take North Carolina or Arizona. He just has to do a hair better than Clinton in Pennsylvania, Wisconsin, and Michigan.
Though Trump’s 304 electoral votes made 2016 look like a solid win — after all, it was better than George W. Bush’s two victories — in reality, the decisive states of Pennsylvania, Wisconsin, and Michigan were all decided by less than 1 point.
So, landslide predictions aside, does data give us reason to believe Biden is going to do a smidgen better than Clinton?
First, let’s start with the national polls. To be clear, yes, there is not a national election, so these polls are not conclusive, but there are a few aspects of national polls I find noteworthy.
To start, national polls were not as far off as state polls in 2016. The final RealClearPolitics average had Clinton up 3.2 points nationally, and she won by 2.1 points. On Tuesday morning, Biden was up 7.2 points nationally in the average. While, again, state-level results are what matter, it would be hard for Biden to improve 4 points nationally over Clinton’s performance without making up at least a point in those three key states. Also, one thing worth looking at is the trend. There were several points in the 2016 race in which Trump was virtually tied or even slightly ahead of Clinton. Yet, Biden has been consistently ahead of Trump for over a year, and Trump has never trailed by less than 4 points.
On the state-level polling, there’s a bit of a divergence between the RealClearPolitics and the FiveThirtyEight averages. So let’s look at both.
For RealClearPolitics, Biden is up 6.7 points in Wisconsin (Clinton was up 6.5 points in the final average); 1.2 points in Pennsylvania (vs. 1.9 Clinton final lead); and 4.2 points in Michigan (vs. 3.4 point Clinton final lead). If the RealClearPolitics polling average were to err by the same percentage this time around as it did last time, that would mean Biden would win Michigan, but that Trump would retain Wisconsin and Pennsylvania. All else being equal, this would be enough to get Trump reelected.
But applying this methodology more broadly (i.e., applying the 2016 polling errors to the final RealClearPolitics 2020 averages) produces some counterintuitive results. That is, if we were to make similar adjustments for all battleground states, Trump would also carry Minnesota, North Carolina, Florida, Georgia, Ohio, and Iowa. However, Biden would carry Nevada, Arizona, and Texas — and win the electoral vote 287 to 251. In practice, it is hard to imagine a scenario in which Trump takes Minnesota and Wisconsin while Biden takes Arizona and Texas.
If one looks at FiveThirtyEight, its final polling averages have Biden up by 4.7 points in Pennsylvania, 8 points in Michigan, and 8.3 points in Pennsylvania. Those leads are substantial enough that they point to a Biden win, even if polls are just as wrong as they were in 2016. Throwing in another data point, FiveThirtyEight crunched the numbers as to what would happen in the states if there were a 3-point national polling error against Trump. It found that Biden would still win, but with a narrow 279 electoral votes.
Basically, the scenario in which Trump wins is one in which polls are once again understating his support in Pennsylvania, Wisconsin, and Michigan — and that he hangs on in the other battleground states.
It’s worth noting that pollsters have insisted that they have corrected for 2016 errors by including more non-college-educated whites in their likely voter models. That would suggest polls could be more accurately accounting for Trump’s support this time, or perhaps even overestimating it.
The earliest sign of how the race is likely to play out is in Florida, where the votes are expected to come in quickly. If Biden is still holding the lead after the votes from the conservative panhandle come in, he will most likely take the state. If Florida is called for Biden, it not only makes the electoral math nearly insurmountable for Trump, but it suggests that we can have more confidence in the other polling pointing toward a Biden win. Because Biden has other paths to 270, a Florida win would not be as devastating a blow to him. But it would at least point to more of a nail biter of a race that likely won’t be decided for days.
