The Economist: Forecasting the US elections

[GeorgiaModerate]

Here's their writeup on the methodology: https://projects.economist.com/us-2020-forecast/president/how-this-works.  Long but interesting.

[GeorgiaModerate]

*yawn* have heard this before. Just wait when we get to November.

These are predictions and forecasts based on *current* polling and info. No one is saying that this is absolutely certain to be the case in November no matter what. Just saying “wait till November” doesn’t negate the current state of the race, which looks terrible for Trump.

Actually, at this point the model is almost entirely based on fundamentals.  For more details on the technical nuances, see this thread (well worth a read if you're interested in such things): https://twitter.com/DanRosenheck/status/1271119100962656256.  In particular, this tweet jumps out:

https://twitter.com/DanRosenheck/status/1271119106545266688

[GeorgiaModerate]

https://twitter.com/gelliottmorris/status/1271291118424997889

[GeorgiaModerate]

Oh great, I guess even more ammo that Republicans and Trump supporters will use if Trump wins reelection. Very nice The Economist.

Especially now it is not the time to claim Trump has only a 15% chance of winning. No matter the polls, as of now it is impossible for him to be below 40-45% chance of winning, it is still June.

I've been thinking about this.  The election is far from over, but it's undeniable that Biden has a significant lead at this point and should be considered the favorite.  Anyone who disputes this is simply not looking at the data objectively.  (Note: responses citing subjective factors or feelings will be ignored with prejudice.)

One of the model developers noted in a Twitter analysis thread that if they had a "nowcast" showing the result if the election was today, it would have Biden as a >99% favorite.  Or another way of looking at it: if all conditions (poll numbers, economic factors, etc.) are the same on Nov. 2 as they are today, Biden will be an overwhelming favorite to win.

Obviously, the uncertainty in the outcome is due to the time left during which things can change.  If we call the probability of a Biden win "X", then at this point .5 < X < 1.  As things change (new polls, economic changes, and time passing) X may move up or down.  We can all make guesses for X based on our own intuition and experience, but what the Economist team has attempted to do is quantify the various factors that contribute to uncertainty, and thus to the value of X.  Their methodology might be incorrect on some of these factors, but at least they've made an attempt to be rigorous about it.  So for now, I'm willing to believe their estimate of X is a reasonable picture of the state of the race.  It might be a little off, but I don't think it's very much off.

[GeorgiaModerate]

https://twitter.com/gelliottmorris/status/1272636770530856961

[GeorgiaModerate]

https://twitter.com/gelliottmorris/status/1273288676949266435

[GeorgiaModerate]

I am immensely triggered by the fact that they don't have separate numbers/forecasts for Nebraska's and Maine's congressional districts.

It's coming.  Elliott has said it takes a lot of code to properly account for them, and it doesn't change the overall result that much, so they launched before that part of it was implemented.

[GeorgiaModerate]

https://twitter.com/gelliottmorris/status/1273393888166887424

[GeorgiaModerate]

https://twitter.com/gelliottmorris/status/1275162895190458369

[GeorgiaModerate]

When was the last time Trump was at -5 in either RCP or 538?

He got up to about -4 in 538 during his brief coronavirus bump in early April.

[GeorgiaModerate]

Biden hits 90% in the Economist model.  Elliott's comments:

https://twitter.com/gelliottmorris/status/1276581009224540161

[GeorgiaModerate]

https://twitter.com/gelliottmorris/status/1278071285986328578

[GeorgiaModerate]

Interesting that they put Michigan in the same bucket as Nevada, Colorado, Virginia, Minnesota, Maine, and New Hampshire.

Also interesting that they think Biden has a better chance of winning Florida than Trump has of winning Texas.

And South Carolina closer than Michigan!

[GeorgiaModerate]

This is a very cool animation:

https://twitter.com/gelliottmorris/status/1278065258020646914

[GeorgiaModerate]

https://twitter.com/gelliottmorris/status/1283979526868029440

The model currently has Biden as a 93% favorite.

[GeorgiaModerate]

I read the whole (long) discussion.  It was interesting.  But I'd like to see the Nates get their own models published rather than criticizing other ones, although some of their points do seem potentially valid.

[GeorgiaModerate]

Elliott Morris has a Q&A Twitter thread going on right now, taking questions about the model:

https://twitter.com/gelliottmorris/status/1288252792268783617

[GeorgiaModerate]

It appears G. elliott got flack from Dave Wasserman and others and ended up reworking his model. Now he's claiming that Trump had a better week last week (not true, according to the polls), and Biden dropped a whole 2% in his model.

These people need to stop letting other people affect how they do their work

I don't agree with the last part.  This isn't an exact science.  It's perfectly reasonable for other experts in the subject to offer constructive criticism, and I think that's the spirit in which it's been offered and received.  I think it's a good thing if it results in improvements in the model.  The Economist's model is out there before the 538 or NYT models, so it's going to be the first target of such criticism (and the Economist team deserves a ton of credit for being the first ones to publish!)

[GeorgiaModerate]

The map on the model page has now been improved.  You can now hover over any state and see the individual probability for that state.

https://projects.economist.com/us-2020-forecast/president

[GeorgiaModerate]

Silver just sounds more and more pissy.

As others have said, Nate's position in this argument would be stronger if he'd just publish his own model.

[GeorgiaModerate]

https://twitter.com/gelliottmorris/status/1289198771780239364

The linked article is worth reading.

[GeorgiaModerate]

https://twitter.com/gelliottmorris/status/1289941988880445440

[GeorgiaModerate]

Interesting:

https://twitter.com/gelliottmorris/status/1290339003070324738

[GeorgiaModerate]

https://twitter.com/gelliottmorris/status/1290857816786382849

[GeorgiaModerate]

The new model is up (same URL: https://projects.economist.com/us-2020-forecast/president).

Biden's probability of winning the EV went from 91% to 88%, and of winning the PV from 99% to 97%.  But his median EV went from 347 to 351.  Also, Texas has now moved into their Tossup category.  Basically, his position looks slightly stronger in the tweaked model but with the percentages somewhat lower to account for uncertainty (which I think is fair, given the amount of time until the election).

ETA: more details:

We fixed a series of medium-sized bugs in the model that collectively caused too little uncertainty in our forecast. The new model better represents the lower confidence we have both in our fundamentals-based “prior” prediction, and in the polls when we are further out from election day. Statistically, the change is three-fold. First, we increased the band of uncertainty around the prior for every point in the election year, creating a minimum standard deviation of three percentage points on the Democratic vote share nationally. The previous model had a minimum closer to one-and-a-half points. Second, we modestly increased the amount of movement our model allows in the polls across the whole cycle. The average national vote share now moves with a standard deviation of five percentage points over the course of the entire election—up from about 4.5 in the previous model. Finally, we performed various minor tweaks to how we judge the correlations between the states, how we update state-level trends with national data and how we calculate the bias in pollsters that don’t weight their data to match the partisan composition of the electorate. In sum, these changes caused Joe Biden’s chances of winning the popular vote and electoral college to each decrease by roughly three percentage points.