Election 2020: Day -4

I’ll start with a couple of pieces of Minnesota-related news today, which tends to catch my eye since it’s the state in which my wife is registered to vote.

First, this afternoon both Trump and Biden are holding events in Minnesota, Biden in St. Paul (a drive-in rally at the state fairgrounds) and Trump in Rochester (with a 250-person attendance limit due to current state limitations on public gatherings). I’d commented previously that Minnesota is one of the few states where Trump is on the offense, although polling and modeling suggests that by now Trump’s chances of winning the state are somewhere between slim (FiveThirtyEight thinks 7%) and none (The Economist thinks <1%).

That makes me wonder, what is Biden seeing that makes him want to invest the time to stop in Minnesota, even if it is conveniently wedged between other Upper Midwest stops today? The Senate race has been not been viewed as particularly competitive, even before Republican challenger Jason Lewis’ recent health issues. Perhaps he’s trying to help boost Democrats’ chances of flipping the Minnesota state senate, in the hopes of giving the party total control of the state going into a post-census redistricting year? (Right now Minnesota is an increasingly-rare example of a state where the Governor and House are in one party’s hands while the Senate is controlled by the other party.) It’s a bit of a head-scratcher.

Second, last night the 8th Circuit ruled that any Minnesota mail-in ballots received after 8pm on election day will need to be set aside, in order to preserve the ability for later litigation to determine whether or not they should be counted. This is part of a theme of Republican election-related litigation across multiple states, arguing that changes to election rules by state administrative officials that were not ratified by state legislatures violate Article II Section 1 of the U.S. Constitution. (As such, it seems possible that these ballots might eventually get counted with respect to state races, but not with respect to federal races). There is a fascinating legal question here about what the meaning of the word “Legislature” in the U.S. Constitution is; however it is regrettable that this legal argument has only become so visible mere days before an election that is generally acknowledged to be exceptionally important.

In non-Minnesota news, early voting levels in Texas have already exceeded the total number of ballots cast in 2016, with today being the last day for early voting under Texas law. This sounds astonishing. However, I was surprised to learn that almost three-quarters of the total votes cast in Texas in 2016 came through early voting.

In yesterday’s post I threw around some numbers from FiveThirtyEight‘s election models. There’s another widely publicized probabilistic election model floating around, from The Economist. I thought it would be interesting to compare and contrast how these two models are viewing the key races with 4 days to go. As such I pulled some data from both websites within minutes of each other, mid-afternoon on day -4, and here’s what I found.

President. Economist is more bullish than 538 about Biden’s overall chances, with Biden at 96% to win, versus 90% now (up from 88% yesterday) for 538.

I think a major driver of the difference is how the two models view Pennsylvania. While 538 has Biden’s chances in PA at 86%, Economist has Biden at 94%. And as discussed in yesterday’s post, Pennsylvania appears to be a very critical state.

A related driver is that 538 seems to see greater chances of Trump prevailing in several other states that both models view as strong Biden states, like New Hampshire (Economist 98%, 538 89%), Minnesota (Economist >99%, 538 93%), Nevada (Economist 94%, 538 90%), Michigan (Economist 98%, 538 96%), and Wisconsin (Economist 97%, 538 94%).

When it comes to the true battleground states, the two models’ views are fairly well-aligned, but again there’s some tendency for the Economist’s model to compress probabilities towards the likelier outcome. All numbers here are expressed in terms of Biden’s probability of winning:

  • Florida: Economist 77%, 538 66%
  • Arizona: Economist 76%, 538 69%
  • North Carolina: Economist 70%, 538 66%
  • Georgia: Economist 58%, 538 58%
  • Iowa: Economist 45%, 538 46%
  • Ohio: Economist 38%, 538 45%
  • Texas: Economist 26%, 538 35%

Unfortunately the Economist’s model does not appear to produce a forecast for Nebraska’s 2nd District, which as discussed in yesterday’s post could possibly be important in some scenarios; the 538 model currently has this as Biden 78%.

Senate. Similarly, The Economist is a little more bullish than FiveThirtyEight about the Democrats’ chances of prevailing.

There are several races where there’s general agreement about what’s going to happen, but the 538 model is less certain about the outcome than the Economist model. All numbers here are expressed in terms of the probability of the seat flipping parties:

  • Alabama (D-Jones): Economist 99%, 538 81%
  • Colorado (R-Gardner): Economist 95%, 538 85%
  • Arizona (R-McSally): Economist 88%, 538 81%
  • Mississippi (R-Hyde-Smith): Economist 9%, 538 14%
  • Michigan (D-Peters): Economist 4%, 538 17%
  • Minnesota (D-Smith): Economist 2%, 538 7%

On the other hand the two models are aligned with respect to New Mexico’s open Democratic seat, as well as McConnell’s seat in Kentucky, thinking there’s only a 3% to 5% chance either of those seats will flip.

For most of the seats that are in greater doubt, in general The Economist’s model likes the Democrats’ chances better than FiveThirtyEight’s model. Again all numbers below are expressed in terms of the probability of the seat flipping parties:

  • North Carolina (R-Tillis): Economist 72%, 538 63%
  • Maine (R-Collins): Economist 71%, 538 59%
  • Georgia (R-Perdue): Economist 60%, 538 42%
  • Iowa (R-Ernst): Economist 57%, 538 54%
  • South Carolina (R-Graham): Economist 35%, 538 23%
  • Texas (R-Cornyn): Economist 20%, 538 14%

One exception here is Montana (R-Daines), where both models have the seat at 34% to flip. And then there’s the second seat in Georgia, Loeffler’s seat. As I mentioned yesterday, this “jungle general” seat appears destined to go to a runoff, between Warnock (D) and either Loeffler (R) or Collins (R). As such, trying to forecast what’s going to happen here seems particularly difficult, as you’d need to forecast voter enthusiasm for the runoff election, which could depend on how important the outcome is or isn’t with respect to the total Senate picture. Both models have this seat in the 60-63% range to flip, but I would take that with even more of a grain of salt than the other predictions.

All told, the two models generate fairly similar aggregate conclusions: The Economist thinks the Democrats are at 82% to recapture Senate control, while FiveThirtyEight has the Democrats at 77%.