Explaining Trump's Travel Schedule: What Andrew is Watching 6 Days out From Election Day (Patreon)

Remember to join us for our live Q&A and stream on Election Day, Tuesday Nov. 3, from 8-10 pm EST!

If you haven't been to 270towin.com, that's the best place to go to play around with electoral vote maps.  Here's the good news up front:  if the state polls are largely correct, Joe Biden will win handily, perhaps even on Election Night itself.  But -- as we all know painfully well from 2016 -- sometimes, state polls are off systematically.  Our best polling averages on Election Night 2016 showed Clinton leading by 3.7% in Pennsylvania, 4.2% in Michigan and 5.3% in Wisconsin.  You already know Trump pulled off narrow wins in those three states to eke out victory in the electoral college.

So -- and maybe this is Thomas's Negatron wearing off on me -- I'm asking myself two questions:  1) what would happen in 2020 if there was a similar systematic polling error in Trump's favor?; and 2) Is there actual evidence of such an error?

That second question is really important, and let me unpack it a bit.  A systematic polling error is a mismatch between the pollsters' sampling methodology and the actual voting pool.  Keep in mind that campaigns and news outlets pay pollsters to get this stuff right, so they have very strong incentives to adjust their methodology to make sure they don't make the same mistakes twice.  In particular, there's an in-house debate among pollsters as to whether the major systematic error in 2016 was a failure to weight properly by education, or whether education was just one of multiple factors.  And remember that systematic polling errors can run in either direction!  It's possible the corrected models are underestimating Biden's strength and overestimating Trump's; if that's the case, then states like South Carolina, Alaska, Missouri, and Montana could be in play (!).

But let's talk about the worst-case scenario:  what if 2020 is like 2016?  The one I've attached is mine which reflects the current state of polling (as of 8:30 am on Wednesday, October 28) assuming a normal polling error for Trump.  If that's the case, Michigan (current average:  Biden +8.2%), Wisconsin (Biden +8.3%), and Minnesota (Biden +8.5%) would still be very light blue -- but Pennsylvania and Nevada would suddenly be in play.

In that situation, Biden is still strongly favored, but Trump has a plausible path to victory that runs through Pennsylvania and Arizona -- which is why the Trump campaign made multiple stops in Pennsylvania over the weekend and has a similar tour through Arizona today.  

Are there reasons to unleash your inner Negatron?  Maybe.

If you look at the recent state of polling in Pennsylvania, it looks like a solid and perhaps slightly-widening Biden lead of 5 to 7 points, but there are outliers like the 10/26 Insider Advantage poll showing Trump with a +3 lead after their likely voter screen.

What I look for in evaluating outliers is movement.  So I scroll down to find the last Insider Advantage poll, which was 11 days earlier (10/15).  That poll showed a Biden +3 lead -- which, again, was smaller than comparable polls at the time.  So it's possible that Insider Advantage's likely voter screen is Republican-leaning, but even if you think that's the case (I do), they're still showing a 6-point move from Biden to Trump over the past two weeks.

A brief interlude from Optimist Prime here:  other pollsters in Pennsylvania seem to be showing movement away from Trump.  Ipsos, for example, had Biden leading by +4 on 10/19; that lead is up to 7 points in their latest survey.  Trump's barrage of superspreader public events certainly looks like desperation, and there are intuitive reasons to think that these Trump rallies are actually reinforcing Joe Biden's closing argument that you can't re-elect a President who's handled COVID-19 this badly.

So, between now and election day, I'll be monitoring the data very closely and updating this map if need be.  And on election day, I'll be looking to see how the actual data on the ground compares to this worst-case scenario.

The first full poll closings come at 7 p.m. Eastern in Georgia, South Carolina, Virginia, and Vermont.  At 7:30, the polls close in North Carolina, Ohio, and West Virginia.

Obviously, Vermont and West Virginia will get called the second the polls close.  But listen carefully for how the other states are characterized during that first hour before the OA livestream starts.  If the polls are (roughly) correct, that means Georgia, North Carolina, and Ohio should all be listed as "too close to call" -- which means exactly what it sounds like.  But if any of those states are categorized as too early to call, it means that that exit polls are showing a decisive lead for one of the candidates but they're waiting to confirm that the actual ballots match up with the polling.  A decisive lead by either candidate in any of those states would indicate a serious polling error.

How can you know which way that error might point?  Well, you could try to read body language -- if the MSNBC hosts are giddy that a state is too early to call, you might infer that's a Biden lean, or vice-versa on Fox.

But there's one more early data point, and it runs in the opposite direction:  if South Carolina is "too close to call" at 7:00 pm and has not been called by the time our stream begins an hour later, that strongly suggests that maybe the polls have underrated Biden and/or that undecided voters have decisively broken for Biden.  Trump leads Biden by an average of 8.3% in South Carolina -- coincidentally, almost exactly the margin that Biden leads in Michigan, Wisconsin and Minnesota.

Obviously, we'll have lots more decisive data as the night continues, particularly between 8 and 10... which is why we picked those times for our livestream.  We'll see you on Tuesday!