If Warnock wins by 2 and Ossoff wins by 1, as the NYT needle now predicts, that would among other things be a much-needed outstanding performance for pollsters.
If you'll indulge some brief horn-tooting, I'm excited to announce the launch of Graphic Detail, a brand new section in the weekly print Economist. It will be dedicated to illuminating, eye-catching visualizations of fresh datasets and original statistical research. (1/2)
The Dems really did underperform their Senate polls--not just in the states they lost. Manchin, Brown, and Stabenow all undershot by 4-7 points. Had HRC won in '16, the GOP might well have gotten 60 this cycle.
THREAD re the "everybody's-got-it" school of covid truthery, which my story this wk in
@TheEconomist
, based on study by
@inschool4life
&
@Alex_Washburne
, supports. Most impt pt is that these results do NOT mean we should end/loosen lockdowns now/soon. 1/n
New: Iowa Sen Joni Ernst raised $7.2 million in Q3 and has $4.3 million COH.
Her opponent Theresa Greenfield previously announced $28.7m Q3 with $9m in the bank
Data on visits to doctors for "influenza-like illness" imply covid-19 has spread 200x faster than official US data show. Counterintuitively, if true, this would be good news. New in
@TheEconomist
1/2
Man, Obama's victory in Indiana in 2008 really was extraordinary. McCain didn't contest it and I think he somehow got help from it being close to his HQ in Chicago?
After a few technical hiccups this morning,
@TheEconomist
's US Congressional forecasts have been finalized. Dems are >99% to hold the House, with an avg of 244 seats, and 80% to flip the Senate, with an avg of 52.2 seats. 1/n
Again, no evidence that the institutional GOP is able to clear the field for DeSantis for a head-to-head race against Trump. He won the 2016 primary thanks to support from a plurality of dedicated supporters against divided opposition...
This week’s .
@TheEconomist
cover package: Why America’s antiquated voting system is the ultimate cause of partisan polarization, legislative gridlock, and minority rule——and how to start fixing it without touching the constitution:
@baseballcrank
No one is talking about “cutting Congress out of the process of funding the federal government.” If Congress chooses not to fund the govt, it causes a shutdown. Raising the debt ceiling is just a necessary consequence of Congress’s prior choices to fund the government.
Bernie Bros were right all along to call Warren a snake. She's mostly respected their nonaggression pact. But one of the biggest reasons Sanders faces a big delegate deficit is Warren's evisceration of Bloomberg at the debate, which let the moderate lane consolidate around Biden.
I'm a white, non-born-again, straight Jewish man, aged 30-44, doesn't go to synagogue, unmarried, no kids, has a college degree, live(d) in a city in the Northeast, & speaks Spanish. The Economist's model says I'm 76% to vote Democratic. Try it yourself at
@NateSilver538
When FL was a 5-pt error in Trump’s favor, OH and IA were like a 6/7-pt error the same way, and WI/MI/PA hadn’t counted much mail yet...is there any way to reconstruct a retroactive minute by minute probability?
@ianbremmer
@TheEconomist
I'm TE's data editor. Why do you consider this such a big mistake? What do you think her odds are? Might I interest you in a friendly wager?
Looks like Ralston's calling NV for Biden based on the early vote (it's the 1 state where such estimates have been reliable in the past). Do
@gelliottmorris
@Nate_Cohn
@NateSilver538
@Redistrict
know whether this means we should raise our expectations for Biden in neighboring AZ?
Early voting blog is updated.
With Clark Dem firewall approaching unprecedented 90K voters after overnight update, Trump's path in NV leads off a mail ballot cliff.
The dice are cast, and they look like snake eyes for the GOP.
Brief post, more later:
Today's 🔥 Dem primary take: Between this YIMBYism and his forthrightness on gun buybacks, notorious lightweight
@BetoORourke
is more serious on policy than wonk-icon
@ewarren
(whose claim to seriousness is undermined by her proposed fracking ban and raging protectionism). 1/2
"Here's a tough thing to talk about, though we must: Rich people are going to have to allow, or be forced to allow, lower income people to live near them"
@BetoORourke
says in Boston adding "What if, as we proposed to do, we invested in housing that is closer to where you work"
Wow. I’ve rolled my eyes at the people hyping and denouncing her as partisan trolls until now, but she really is a Gabbard/Corbyn-level troglodyte on foreign policy and an embarrassment to her party. Someone primary her please.
If we really want to support the Venezuelan people, we can lift the economic sanctions that are inflicting suffering on innocent families, making it harder for them to access food and medicines, and deepening the economic crisis.
We should support dialogue, not a coup!
Based on the vote tallied so far, Reps have actually won MORE House votes than Dems. That's likely to change as slow-counting CA, NY process ballots, and you have to adjust for uncontested districts. But we're poss looking at the biggest generic-ballot polling miss since 1980 1/2
Since there's not a lot of interest or drama surrounding the presidential vote right now,
@CookPolitical
created a National House Vote Tracker to keep you occupied. Check it out: ?
@collprof
@Michael_Cobb68
I would do so if the data merited it. For example, I’d gladly say there’s a >99% chance that the temperature in Iceland will be below 100 degrees F for every single day in November (and would gladly bet money accordingly). We mean what we say. What is wrong with our framing?
The S&P 500 has done well for investors. But according to a proprietary dataset of 1.6m transactions kindly provided by
@winebid
for
@theeconomist
, a cellar of 500 red Burgundy wines would have done even better. h/t
@mktgmann
Yes, in-person voting really did cause an increase in covid-19 cases in the US. New study by
@daniella_raz
of
@TheEconomist
uses difference-in-differences + 22-variable regression to estimate 220K fewer cases if everyone had voted absentee
Biden's at only 55-60% to win in prediction markets, but 85-90% in other statistical models, which seems like an insanely wide spread—and I don't think either number is super easy to defend—so I have to say I'm pretty happy that 538 model's falls somewhere in the middle.
@andrewbhall
There was a quite big polling error in Trump's direction in a lot of key states, and Biden still won. I do think that validates the models showing that Biden's poll leads were big enough, in enough not-so-correlated states, to make him an overwhelming favorite.
AZ Republicans in competitive House seats outran Lake, who outran Masters/Finchem. That's a measure of the D/R gap in candidate quality/strategy. Lake/Masters also underperformed relative to fundamentals: GOP did better in 2022 than in 2020, but they did worse than Trump/McSally.
The outcome in Arizona is a difference of fewer than 21,000 votes. Less than 1 percentage point. A race that close doesn’t prove the wisdom of any candidate’s strategy.
@NateSilver538
Yeah but WI is fully done counting and is a 7-pt miss. Biden just barely dodged a bullet in a state that polls suggested he didn't have to worry about.
@collprof
@Michael_Cobb68
I did read it. The problem with HuffPo (and Wang) giving HRC 99% in the electoral college was that it was a badly specified model. She was only up 3 pts! Presumably you don’t think that one forecaster’s bad model should preclude the publication of (hopefully!) better ones.
Dear God. Manchin is easily the most valuable senator in the Democratic caucus. If he were to get primaried, he'd be replaced by a conservative Republican.
Hey guys, please ignore this type of garbage. The truth is that elections are never decided on election night. In Utah (and most states) it takes 2 weeks to finalize counting and certify results. It really doesn’t matter who is ahead on election night, it only matters when...1/
This is a helpful reminder. George W. Bush was also up by 10-11 pts on Gore in August before going on to lose the popular vote. However, in my view, the VERY big diff b/w 2020 and these prior cycles is fundamentals. Both 1988 and 2000 featured parties with popular 2-term 1/n
@benjaminwittes
One reason is that Syria had a long-running civil war in which Russia was a late entrant, as opposed to starting a new war by invading a peaceful country and seeking to conquer it and topple its government.
Breaking News: Joe Biden won Arizona, holding a narrow lead there after over a week of counting. The state hadn’t voted for a Democrat for president since 1996.
@collprof
@Michael_Cobb68
The NYT/
@Nate_Cohn
election-night needle? That's the best live-updating probabilistic US election forecast that I'm aware of. I regard it as an outstandingly informative product, one I'd be proud to replicate. What "asterisk/explainer/caution tape" does the NYT needle lack?
PS I'll note as well that what may turn out to be a pretty catastrophic miss in the lower chamber is entirely my fault.
@gelliottmorris
only built our POTUS model, which did extremely well despite the big polling errors, and had nothing to do with this disappointment.
It implies covid-19 might have a similar death rate to <gasp> the flu, but is FAR more contagious. The reason NY hospitals are swamped would be that it's equivalent to cramming a year of flu cases into 1 week. Paper by
@inschool4life
&
@Alex_Washburne
.
2/2
@Nate_Cohn
@collprof
@Michael_Cobb68
That both polls and fundamentals point to Trump losing by 7-11 pts with 3.5 months to go. For Trump to win the pop vote, they would both need to be wrong in the same direction—by a far greater amount than they jointly have been in the history of postwar presidential elections.
@trivialscholar
@NateSilver538
No, you can't. If your 65% events wind up happening 90% of the time, your model is awful. They should wind up happening, give or take...65 times out of 100.
@collprof
@Michael_Cobb68
This is a forecast, not a now-cast. It reflects what’s likely to happen in Nov, not now. For Biden to lose the popular vote, mid-July polls would have to be off by 10 points That can happen—it did in 2000—but if it does, it’s generally in the direction of the fundamentals. 1/2
I believe GA is unusually easy to poll, b/c the non-col white vote is so GOP & the state has so many black voters that demographic weighting gets u a lot of the way. "Swing voters" are college whites, who like talking to pollsters and have high, predictable turnout.
@Nate_cohn
?
@nntaleb
@NateSilver538
Wait, on what planet does Nate not get that principle? He writes all the time about the calibration performance of 538's models...
What was the last time the president's party flipped even one state legislative chamber in a midterm? Maybe there's a recent example that doesn't occur to me...
BREAKING: Democrats flip both chambers of the Michigan Legislature, will have a trifecta in Lansing for the first time in 40 years with control of the House, Senate and Gov.
@GretchenWhitmer
.
#ElectionDay
#Election2022
This *is* bizarre. 3 super partisan R polls and no NYT/Siena with mega sample size. I don't suspect malice, but do suspect negligence. (h/t Hanlon's Razor)
the RCP average for PA now entirely excludes this weekend's 1800 interview NYT/Siena poll, for some reason.
(Trafalgar finished fieldwork the same day or earlier and is included, along with Insider Advantage and Susquehanna, two other GOP-aligned pollsters)
Polls of indiv House districts were also consistent w/ D+7 nat'l environment. Dems had v big fundraising edges, & enjoyed incumbency advantage in many districts where they didn't in '18. This will be a v bad yr for my House model. No idea what data I could've used 2 do better 2/2
@ProfHansNoel
Mismatch between voting rules that guarantee a 2-party system (courtesy of the 12th Amendment and Duverger) and checks and balances on lawmaking with lots of veto points, requiring broad coalitions to get anything done.
Reestablishing a baseline level of broadly recognized facts is both essential and seemingly impossible. There will always be a farther-fringe source eager to serve up fabrications to people who don’t want to be told they were wrong.
PSA regarding this gloriously ratioed tweet. Perhaps ill-advisedly, I combined a sarcastic point with a serious one. For the record, I do not actually think that Elizabeth Warren is a 🐍, literally or metaphorically. In fact, I don't think she has done anything wrong at all. 1/10
Bernie Bros were right all along to call Warren a snake. She's mostly respected their nonaggression pact. But one of the biggest reasons Sanders faces a big delegate deficit is Warren's evisceration of Bloomberg at the debate, which let the moderate lane consolidate around Biden.
Technical THREAD on the
@theeconomist
's new presidential forecast, by
@gelliottmorris
w/ help from me,
@StatModeling
and
@MHeidemanns
. First, Biden's 82% topline is high. Uncomfortably high. Intuitively, it doesn’t feel right to have anyone this heavily favored this far out 1/n
Terrific
@ronanfarrow
deep dive on NSO Group, with access to CEO & staff & great play-by-play of their cat-and-mouse game with WhatsApp. There’s even a Rickroll!
@collprof
@Michael_Cobb68
And we are indeed writing for a large, non-specialist audience. Our published articles lay out in great detail how our models work, and why we believe their stated probabilities are accurate. How do you fear we are misleading our readers? I’m very open to constructive criticism.
@PatrickRuffini
Trump was the only primary candidate to offer exactly what they wanted--keeping POC and women in their place, and protecting govt handouts for old white people (SS and Medicare). That genie is now out of the bottle, and can't be put back. No Ryanist candidate can compete with it.
@LPDonovan
Actually our model shows 15% that Trump wins at least 1 of MN/NH/NV/ME. I was surprised to see it was that high, but correlations aren’t 1. 538 has a 25% chance that Trump wins a Clinton state (sounds too high to me).
Bloomberg has surged to 13% on
@predictit
to be the Dem nominee: . Seems v high to me. He's taken evenly from Warren/Biden/Buttigieg; Bernie's price is unchanged. 1/2
@PatrickRuffini
Not an accident. The GOP's problem is that very few voters actually want to cut rich people's taxes and cut govt spending on programs they think benefit them. What its voters, many of whom used to be Democrats, want is to protect/regain their status in the American pecking order.
Rice has just overtaken Harris on the
@predictit
betting market for Biden's running mate. That DNC schedule that listed Harris's speech separately from "THE VICE PRESIDENTIAL NOMINEE"'s, and did not name Rice anywhere else, might be why.
@collprof
@Michael_Cobb68
@Nate_Cohn
The NYT's models are in my opinion very well-calibrated. If they say 99%, it really is 99% (unlike Wang/HuffPo 2016). I completely agree that conveying uncertainty accurately is important. But that means emphasizing when uncertainty is low as well as when it is high, no?
Golf stats fans: I'm excited to announce that
@TheEconomist
will launch a new, live-updating, interactive visualization of our
#EAGLE
statistical golf prediction system for the
#Masters
next week. 1/n
I rarely tweet about my team's work, but this magnum opus from
@Sondreus
is worth an exception. You can explore
@TheEconomist
's estimates of covid-19 death & hospitalization rates for 80 billion combinations of age/sex/comorbidities at
@FrankLuntz
Yes but the year-on-year rate is going to keep looking high until enough time passes that the trailing 12 months no longer includes the super-high-inflation months earlier this year. The past few months of data look much more contained.
Very interesting to see a poll in a deep red and not very elastic state. Trump won there in 2016 by 26, so this is a 17-point margin swing from Clinton to Biden—much larger than the shifts we're seeing in more competitive states.
@gelliottmorris
The lovely Amy McGrath is a Marine Corps veteran. The average donation to her campaign is $36.
Mitch McConnell is a broken-down old man owed by cheap labor lobbyists.
#DefeatMcConnell
Television viewers not checking Twitter, quant models or betting markets still have absolutely no idea that Jon Ossoff pretty much has this in the bag. This is really journalistic malpractice. It reminds me of the online-TV gap in baseball journalism from like 10-15 years ago.
Republic congresswoman Nancy Mace on the BBC just now: "This is disgusting...I'm a huge supporter of President Trump, supported him for 4 years, but all that was wiped out today." Interesting to see how many GOPers will be so unequivocal.
@Nate_Cohn
@collprof
@Michael_Cobb68
baked in extra uncertainty b/c economic conditions are so far from the norm, & still thinks the notion of an incumbent president running amid mass unemployment making up a 10-pt polling deficit in 3.5 mos to win the popular vote is outlandish. Not impossible, but very improbable.
@Nate_Cohn
@collprof
@Michael_Cobb68
Could this time be different? Absolutely. There are lots of factors our model doesn’t incorporate. Maybe a covid cure will be invented next month, and the economy roars back. Maybe fear of the virus crushes Dem turnout while leaving Rep turnout unscathed. But our model already
YouGov is already out with a SCOTUS poll:
Voters think 51–42% that Trump should not appoint a new justice before 2021.
If he does anyway, voters think 48–45% that the Senate should not confirm.
Props to
@Nate_Cohn
here--the
@UpshotNYT
needle never wavered, keeping Tester's win probability at 80%-plus even when he trailed in the tallied vote by more than a full point.
@DavidInglesTV
Well sure, undecided or 3rd party voters tend to come home to both major parties as the race draws to a close. The only difference is that Biden is at like 52 rather than Hillary’s 46-47. Other than that, Mrs. Lincoln...
@EricTopol
Was some of that luck just that SARS-CoV-2 turned out to be a fairly easy virus to vaccinate against? We've been trying to develop one against HIV for 40 years to no avail...is that because the spike protein is an unusually juicy target for antibodies?
This is embarrassing. 1st of all, we have no idea what the final margins will be in many of these races. 2nd, lots of these are no longer “swing districts” (VA10 lol). Finally, no controls for anything. FWIW I find v little impact for ideology, but more moderate = slightly better
New data analysis from
@justicedems
and
@sunrisemvmt
shows vote share for House Democrats actually declined in swing districts as candidates adopted more conservative views.
Personally, I think Warren (and every other candidate) should do whatever she thinks will help her campaign or the causes she believes in, regardless of its effect on rival candidates she is trying to beat. But that's just me. Apologies again for inadvertently giving offense. END
@Eric_Dykstra
@nntaleb
@NateSilver538
Not sure what you mean by "incorporate noise." There's a reason they play real sports games and hold real elections--because oftentimes the favorite doesn't win. That unexplained/unexplainable variance is the "noise". A good model correctly measures it to produce a probability.
Annnndddd we're off! The first holes of live data have arrived from the
#Masters
. Watch every player's win probability update in real time with
#EAGLE
by
@TheEconomist
. (Win-prob chart will just show straight lines until something interesting happens.)
Yup. Still very far to the right of the country. FWIW AZ’s partisan lean this year will prob be similar to 2018 and 2016 (~5 pt to the right of the country). GA did move 2 pts left.
@BetoORourke
@ewarren
Clearly zoning and land-use are primarily local issues, not federal ones, and any mandatory gun-buyback plan would have trouble with a right-wing SCOTUS. But at least Beto has identified the right direction of travel. 2/2