Monthly Archives: October 2020

Four more years? (one week to go edition!)

2 Replies

Well glory be, there’s only a week to go! Did you ever think we’d make it this far? Me either. But here we are! Biden’s position was looking pretty strong a week ago. What about now? There are several ways we can approach this question…

There being a week to go, we might look back a week, and see if there’s evidence of movement in any particular direction. At the national level, RCP’s average has negligible movement over the past week:

The tiny bit of tightening seen above is frankly indistinguishable from statistical noise. it certainly doesn’t show evidence of a big move for or away from either candidate.

As we’ve mentioned before, RCP has a “naive” average in the sense that it just averages together recent polls. Over at 538, their averages also do weighting by a pollster’s track record of accuracy, and take into account the historical partisan lean of various pollsters. This is certainly more sophisticated, and, arguably, leads to better results. Using this method, their “topline” number for Biden is higher:

They are also showing some tightening, in that Biden’s lead is down from 10.3% a week ago. While this is more movement than RCP shows, it’s equally hard to distinguish from statistical noise within the margin of error, and certainly doesn’t seem like a major shift.

But, as you may have heard once or twice, elections aren’t determined by national results. They’re determined by the electoral college, and how the candidates perform in the individual states. Here’s how the numbers for the most likely swing states have shifted over the past week:

There’s definite evidence of tightening up here, but Biden still leads in all six states with a week to go. Some of these leads are pretty narrow and well within a margin of error, so we wouldn’t be surprised if, for example, he didn’t end up carrying Florida or North Carolina. Crucially, however, RCP has him above 5% in Michigan and Wisconsin, which means any one of the other states could take him over the top. 538 has him above 5% in those two, and Pennsylvania as well, which wins the race.

If we look at RCP’s “no tossups” map today (right) versus one week ago (left) the only change is that they now have the very close Georgia going to Trump:

To sum up here, looking over the past week we see some evidence of tightening (actually a very usual occurrence toward the end of a campaign), but we don’t see any indication of a big trajectory change. And, since Trump is behind by 7-10 nationally and trailing in the most crucial swing states, he needs a trajectory change. “Staying roughly the same” for Biden, is remaining in a predominant position.

This is even more true given that, thanks to the push for absentee and early voting due to the pandemic, literally tens of millions of people have voted over the last week. As of this writing, more than 68 million ballots have already been cast, representing almost 50% of the total votes from 2016:

So the preceding polling numbers weren’t only a preview of what voters may do on Election Day, they are the data that came in as voters were actually voting all across the country.

Another way we might approach this is to look at Biden compared to Clinton at the same point in 2016:

Looking at these side by side, several things become apparent:

  • Biden’s lead is about twice as big as Clinton’s was a week out.
  • Trump’s late surge in 2016 is readily apparent. Nothing like that seems to be going on in 2020.
  • There are considerably fewer undecideds at this point than there were in 2016.
  • Biden remains solidly above 50%, while Clinton was falling down from 48%.
  • Trump hasn’t had a day over 44% since March, whereas at this point in 2016 he was climbing up toward 46%.

Two other things are worth quickly mentioning about the state by state outlook in 2016 versus 2020. The first is that there was a significant third party presence in 2016, whereas we have no indication of it being above the more typical 1%-2% this time. That created a lot more uncertainty in races that were down to the wire in 2016 than there is this time.

The second is that the polling of at least the Midwestern swing states is better this time- both in terms of number of polls, and models taking into account the kinds of voters that went for Trump in 2016. Pollsters may be making entirely different kinds of mistakes this go-around, but being way off in Michigan, Pennsylvania, and Wisconsin probably isn’t one of them.

One can see the net effect of the lower national margin and the uncertainty in 2016 versus 2020 quite clearly in the 538 forecast model. As of today, it has Biden at an 88% chance of winning the race:

At this same point in 2016, Clinton had dipped down to 71% (and was falling):

Any way you look at it, week-to-go 2020 versus week-to-go 2016 indicates that Biden is in a much more solid position than Clinton was. And recall, Clinton won the popular vote, and only isn’t President today because of literally 77,000 voters spread across three mid-western states. It doesn’t take much marginal improvement in 2020 vs. 2016 to swing the election, and Biden looks to have more than marginal improvement.

The last area we might quickly check in on is whether other data points line up with the notion that Biden is doing well. With a week to go, Trump remains around 10% underwater in net popularity, a position no recent President has won from:

538’s Senate model indicates that Democrats remain favored to win back the Senate:

As does RCP’s “no tossups” Senate map:

And the generic Congressional ballot shows Democrats with a clear edge nationally:

As the 538 model gives them a 96% chance of keeping the House, and even expects them to widen their margin by a few seats:

Whether we look at movement from a week ago, comparison to 2016, or fit with other data points, the verdict is clear: Biden remains in an extremely strong position with one week to go.

We’ll do one more update on Monday with (egads!) 24 hours to go!

Four more years? (two weeks to go edition!)

1 Reply

Buddha bless us and Saints preserve us, two weeks to go! At a month to go, my read of prevailing evidence was that Biden was in a very strong position. What about now?

Let’s start with national polling. RCP’s average currently has Biden up by 8.6%:

538, which takes a more nuanced approach to their average, taking into account the accuracy track record and historical partisan lean of individual pollsters, has Biden leading by even more:

A couple of other national “looks” are worth discussing. It being two weeks to go, we might look back over the past two weeks, and see if there’s evidence of major changes that one might project forward to election day. The short answer is “no”:

To the extent that Biden seems to be solidly ahead, every day with no significant change benefits him. Especially given a factor that we’ll discuss in a moment… Another relevant “look” is comparing where Biden is with two weeks to go with where Clinton was at the same point in 2016:

The same factors we’ve seen all year show up here: Biden is ahead by more than Clinton was at the same point, has had a higher “ceiling” and better “floor” than her numbers, and has never led Trump by less than 4 points, whereas Clinton was frequently within 1%-2%, and twice even trailed Trump.

One additional factor to consider is that who’s been ahead in the polls is not an academic matter. In fact, the election is already 24% over. That is to to say, between absentee ballots and early voting, 33.3 million votes have already been cast, representing 24% of the total from 2016:

One of the reasons Comey’s October surprise in 2016 was so damaging to Clinton was that there were a lot of late deciders on the fence. There are far fewer this time, and almost 1 in 4 votes is already in. And these votes came in during a period when Biden was strongly leading. So even if something damaging to Biden (or great for Trump) arose, it inherently can’t have as much of an effect now as it did in 2016.

National polls and vote totals, especially compared to 2016, do give us an idea of how things are trending. But, ultimately, it all comes down to the states. According to RCP’s latest tallies of the most likely swing states, Biden is leading in all of them:

538’s more nuanced approach shows the following Biden margins for these states:

If we were to take the lower RCP numbers and only give Biden anything above 5% and assume the others are a toss-up, this is the map we’d get:

In this scenario, any one of Florida, North Carolina, or Pennsylvania takes Biden over the top, as would Arizona plus the split electoral vote from Nebraska. Trump, conversely, pretty much needs a full sweep to prevail. And this is before getting around to states that Trump won fairly handily in 2016, but which are toss-ups this time- Georgia, Iowa, Ohio, Texas, and Maine’s split electoral vote.

RCP’s “no toss ups” map showing all current Biden leads looks like this:

Based on these state by state polling outlooks, 538’s model has Biden favored to win:

And the Cook Report and Sabato models both have the “lean Democratic” states alone totaling more than the needed 270 electoral votes:

“But the polls in 2016 were wr-” Okay. I know you all are too smart and too well informed at this point to pull out that tired old chestnut and re-roast it. But just in case, here’s a good discussion of what pollsters have done since 2016 to improve, and what areas they’re still worried about. And there are other data points we can examine…

To the extent that fundraising is another indicator of enthusiasm, the signs for the Democrats are very promising. Biden has been out-raising Trump by more than 50%, Act Blue is more than doubling the take of Republican activist platforms, and the Democratic candidates in the leading Senate races are all solidly out-raising their Republican opponents.

Speaking of those Senate races, they form an independent data point of their own. 538 currently has Democrats favored to recapture the Senate:

And the RCP “no toss ups” Senate map is showing the same thing:

Over on the House side, the generic Congressional ballot shows Democrats leading by 7.4%:

This is actually up about a percentage point from a month ago, and indicates the party is in no danger of losing the House. Indeed, the 538 model calls for it to slightly improve upon the 232 seats it currently holds:

One final thing to check in on is Presidential approval ratings. It hardly needs to be said, but, for the record, an incumbent being more than 10 points underwater with two weeks left to go isn’t a great sign for their re-election:

Looking back over the past 50 years of Presidential approval at this same point drives the point home:

Trump’s numbers look nothing like Presidents who were heading to easy victories. They don’t even look like Presidents who were heading to narrow victories. Instead, they most resemble Carter and Bush 1 when they were both on the edge of first-term defeat, and Johnson when he was so unpopular from Vietnam that he had decided not to run again.

To sum up: No matter what indicators you look at, with two weeks to go, Biden seems to be in a very strong position. Given the relative stability of this race throughout the year, I would be surprised if this changes drastically over the next week. But let’s meet back here in a week and see!

Four more years? (one month to go edition!)

1 Reply

It’s hard to believe. Even more so since we’ve had, like, a year’s worth of news in the last week. But, as of this past weekend, it was officially one month to go until Election Day! When we last checked in a month ago, the conclusion was, “Biden remains in a very strong position with two months to go.” If that was true then, it’s even more so now.

Let’s start with national polling. Using their average of recent polls (also note, the majority of these now include post tax story/debate news, but don’t yet have data since revelations of Trump’s COVID diagnosis), RCP currently has Biden leading by 8.3%:

RCP’s average is “naive” in the sense that it just adds together recent polls and takes their average. Over at FiveThirtyEight, they also adjust their weightings for pollster’s history of accuracy, compare national polls to state tallies for a “sanity check”, and take into account pollsters that have a historical partisan lean in one direction or another. This is certainly a more robust methodology, but at the moment it produces a nearly identical result, Biden +8.2%:

There are two things worth noting about this lead. The first is that it isn’t a meaningless data point, since more than 2 million votes have already been cast, and voting is taking place in more than 30 states right now while Biden has this healthy lead:

The other point worth talking about is margin of error. Even the best polling, irreducibly, has a roughly +/- 3% margin of error. To take the above 538.com numbers as an example, Trump could be as high as 45.7%, and Biden could be as low as 47.9%. That means that, at worst (from his point of view) Biden’s lead could be as little as 2.2%. Of course, equally, Biden might be as high as 53.9%, and Trump might be as low as 39.7%.

This matters because, as 2016 showed us, a narrow popular vote lead can still be an electoral college loss, depending on how the individual states fall. FiveThirtyEight has tried to quantify this, and at a 2% to 3% national lead, Biden has around a 50% chance to win the Electoral College. Of course, this currently looks to be Trump’s absolute best case scenario, and if Biden is ahead by more than 3%, his chances shoot up accordingly.

Did I mention individual states above? Indeed, and that is what the election ultimately swings on. As it happens, Biden currently enjoys a lead in all of the most probable swing states:

Granted that some of these leads are narrow, even restricting the map to just states where his lead exceeds 5% takes him over the top:

If we were to go further and look at RCP’s “No Toss-ups” map of all the states based on current polling averages, we get this map:

Using their sophisticated state-by-state model, 538 is currently predicting an 81% chance that Biden wins:

Note that this model in part reflects the uncertainties of still being a month out. To put this another way, every day we get closer to the election with Biden still having these kind of national and state numbers, that percentage rises. For reference, in 2016 538’s model had a 29% chance that Trump would beat Clinton on the eve of the election. Biden is in substantially better shape than that, as a quick side-by-side comparison confirms:

I do this list a lot, but because I think it’s instructive I’ll do it again!

  • Clinton frequently had a lead as low as 2% or less over Trump, Biden has never had a single day under 4%.
  • At two points, Trump actually led Clinton, he has never led Biden for even a single day.
  • Biden has spent significant time over 50%, and almost no time under 48%, whereas Clinton was often below 48%, and only topped 50% for a single day.

National polls, state polls, and the comparison to 2016 all confirm the notion that Biden is in very good shape with a month to go. What do other data points have to say?

According to FiveThirtyEight’s average, Trump’s net approval rating is currently more than eight points underwater:

To put that in historical perspective, Trump’s net approval numbers are worse at this point than any other President of the last 50 years except for Bush I and Carter, both of whom were about to be defeated:

Democrats are leading the Generic Congressional Approval ballot by 6.5%, meaning that their hold on the House is not in any danger, and could even expand slightly:

RCP’s “No Toss-ups” Senate polling averages map shows Democrats re-taking control of the Senate:

And FiveThirtyEight’s Senate forecast model gives Democrats a 65% chance of doing so:

In short, reference to Trump’s net approval numbers and the House and Senate outlook further bolsters the idea that Biden’s lead is as robust as it appears.

There are, of course, factors that this analysis doesn’t take into account. Election night returns may temporarily show Trump leading before absentee ballots are tallied and reflect the full picture. And Trump has given every indication that he plans to dispute results. But none of that, ultimately, changes the fundamental fact that, a month out from the election, Biden has a very solid chance of prevailing on 11/3/20 (plus a week or two).

Tune in again two weeks from now for the two week outlook!