538 starts forecasting the US 2018 midterms.

August 17, 2018 9:07 AM

Classic: I’ll take the polls, plus all the "fundamentals": fundraising, past voting in the district, historical trends and more."

538's new forecasting models for the 2018 House elections are out.

Nate Silver: How FiveThirtyEight’s House Model Works

Nathaniel Rakich: 25 Districts That Could Decide The House In 2018

Nate Silver & Galen Druke: Politics Podcast: Let’s Talk About The 2018 House Forecast

Nate Silver: The 5 Big Takeaways From Our House Forecast

Some excerpts from the "Big Takeaways" article:

...There are a couple of exceptions — indicators that are a little out of the consensus — but both of them fall on the better-for-Democrats side of the consensus. First, Democrats have done really impressively in fundraising...Meanwhile, the results of special elections have been very good for Democrats. Our model doesn’t actually use special election results in its forecasts, but they’re part of a coherent alternative narrative in which there’s upside for Democrats relative to what our forecast shows. Donating money and voting in special elections are tangible indicators of voter engagement, and it’s possible that they point toward a Democratic enthusiasm advantage that could become clearer later on in the cycle.

...

We have only 215 seats rated as favoring Democrats — “lean Democrat” or stronger — which is fewer than the 218 they need to take the House.

Nonetheless, Democrats are favored to win the majority if current conditions hold because they’ll have a bunch of opportunities, even as underdogs, to win those extra seats: 14 toss-up races, 19 “lean Republican” races and 53 “likely Republican” contests. Those are a lot of lottery tickets to punch, even if Democrats aren’t necessarily favored in any individual race.

But Democrats would have a problem if there’s a shift in the national climate toward Republicans, or even if there’s a relatively modest systematic polling error in the GOP’s favor on Election Day. All of the sudden, they’d lose most of the toss-up races along with some of the “lean Democrat” races...

The flip side to this is that if the political environment gets better for Democrats, their seat gains could pile up at an accelerating rate. There are a plethora of districts that are 10 to 20 points more Republican than the country as a whole, a lot of which were gerrymandered to be “safe” for Republican candidates — but where the gerrymanders could fail in the event of a large enough wave.

...

At earlier points in the cycle, Democrats had appeared to need more like a 7- to 8-point advantage in the national popular vote to be favored to claim the majority of seats. Since then, the Republican edge has been eroded by retirements, by Pennsylvania’s redistricting, and by the relatively weak GOP incumbency advantage...

The models will be updated as polling and other new data comes in. A Senate forecasting model is in the works.

More 538 stuff:

generic ballot

pollster ratings

Meredith Conroy, Mai Nguyen & Nathaniel Rakich's analyses of the 2018 Democratic primary niminations:

We Researched Hundreds Of Races. Here’s Who Democrats Are Nominating.

We Looked At Hundreds Of Endorsements. Here’s Who Democrats Are Listening To.

Please keep in mind guidelines for US political discussion on MetaFilter.

Also: This thread is about the upcoming 2018 US elections, polling and election forecasts. Please take news about Trump, White House leaks, the Manafort trials, etc. to the the latest catch-all thread.

posted by nangar (92 comments total)

I’m a nerd and I find this kind of modeling fascinating. But I also don’t think this should affect actions on the ground before the election: Democrats need to work as hard as they can, and not take encouraging projections or talk of a “Blue Wave” as a reason to let up even a little.

It also doesn’t say much about any given race. In the most recent podcast about the model, Silver pointed out that most House races have little or no polling. 538 uses generic ballots and the polls of districts that exist to build their projections, but at most that helps predict the aggregate result. Individual races can vary a lot.

Anyway, this is still not that encouraging a projection: it gives Republicans a 1 in 4 probability to keep control of the House. That’s not too unlikely.
posted by fencerjimmy at 9:22 AM on August 17, 2018 [12 favorites]

Anyway, this is still not that encouraging a projection: it gives Republicans a 1 in 4 probability to keep control of the House. That’s not too unlikely.

Right around the chance that 538 was giving Trump right up to the election, in fact.
posted by Etrigan at 9:24 AM on August 17, 2018 [26 favorites]

Thank you for this post. I'm conflicted because this is just what my midterm-obsessed self wants, in all the gory details - but it's also an alternative scratch for the itch from the good type of panic of getting involved with races. I agree with fencerjimmy - one in four is cold comfort. Let's work hard and have no regrets Wednesday the 7th!
posted by tarshish bound at 9:25 AM on August 17, 2018 [3 favorites]

I'm sorry, but can I really be the only person who, after 2016, has no interest whatsoever in polling from any source?

Eh, 538 was the only outlet that was even remotely close. Seems worth paying attention to them now.
posted by tobascodagama at 9:26 AM on August 17, 2018 [26 favorites]

I'm sorry, but can I really be the only person who, after 2016, has no interest whatsoever in polling from any source?

It was checking 538 in the weeks leading up to the election that sent me frantically scrambling over to Michigan to do campaign work instead of Ohio where friends were telling me to go, not that it did a lot of good. I think they did a pretty good job of making it clear where the vulnerability was on the Dem's side.
posted by tarshish bound at 9:27 AM on August 17, 2018 [16 favorites]

538 had Hillary at 2 out of 3, and I remember trying to explain to a skeptical friend that those were terrible odds considering...
posted by tarshish bound at 9:30 AM on August 17, 2018 [43 favorites]

538 does a phenomenal job bringing in data from all sorts of places and synthesizing it into one comprehensible, beautiful report. That hexagon map is a thing of beauty. So are the numerical models they're using for their forecasts. Sure it's reasonable to be skeptical of any predictions, but this site is a great start for understanding the 2018 election and figuring out where you might want to focus your time or donations.

If like me you lean Democrat and live in California, we have a lot of particularly exciting close races. Josh Harder in CA-10 Manteca, Katie Hill in CA-25 LA/Ventura, Gil Cisneros in CA-39 LA/Orange County, Katie Porter in CA-45 Orange County, and Harley Rouda in CA-48 Orange County are all running very close races. They could use your help! CA-04 Sierra, CA-07 Sacramento, CA-21 San Joaquin Valley and CA-49 Oceanside are also competitive. I'm also a supporter of Audrey Denney in CA-1 Chico / NE California; she's more of a long shot but she's great. I'd love to have her representing me in Grass Valley.
posted by Nelson at 9:31 AM on August 17, 2018 [9 favorites]

tbh, not gonna click. I promised myself that this time around I'd put all the energy I would normally put reading polls and prognostications and spend that time volunteering to actually make a difference, rather than being a passive observer.
posted by coffee and minarets at 9:31 AM on August 17, 2018 [14 favorites]

After their catastrophic miscalculation with the 2016 election polling, Five Thirty-Eight just needs to go away and stop pretending they're relevant anymore. You drop the ball that badly, you go home.

Huh? Other than the Hillary campaign its self, they were one of the first organizations that started to notice the shift to Trump and started to hedge their bets. They only put Clinton as 2 to 1 favorite unlike most other sites which all had Hillary up somewhere between 9 to 1 and 99 to 1.
posted by jmauro at 9:31 AM on August 17, 2018 [50 favorites]

Five Thirty Eight did a great job in the general in 2016 as they always do, but they were totally wrong in the GOP primary, and I hope that is what the 538 critics in this thread are talking about. Otherwise I will be annoyed.
posted by chrchr at 9:32 AM on August 17, 2018 [3 favorites]

Any chance over 50% is a guarantee, right? That's how statistics work?
posted by tobascodagama at 9:32 AM on August 17, 2018 [39 favorites]

538's (somewhat defensive) post-mortem: The Real Story of 2016
posted by tarshish bound at 9:33 AM on August 17, 2018 [4 favorites]

In Illinois, lend a hand to Lauren Underwood in IL-14, Sean Casten in IL-6, and Betsy Dirksen Londrigen in IL-13, whose opponent's field director was just arrested for harassing Londrigen and others at a fundraiser.
posted by tarshish bound at 9:38 AM on August 17, 2018 [2 favorites]

After their catastrophic miscalculation with the 2016 election polling, Five Thirty-Eight just needs to go away and stop pretending they're relevant anymore. You drop the ball that badly, you go home.

They were more accurate than most. Who do you feel is more reliable?
posted by Edgewise at 9:41 AM on August 17, 2018 [7 favorites]

After their catastrophic miscalculation with the 2016 election polling, Five Thirty-Eight just needs to go away and stop pretending they're relevant anymore. You drop the ball that badly, you go home.

538 was about the only outfit to get 2016 right. They gave Trump a 20% chance of winning right up until the end, most everyone else gave him a 1% chance. The also predicted Trump would get 45% of the popular vote, Clinton 49. It was 46/48.

What ball, exactly, did they drop?
posted by Frayed Knot at 9:41 AM on August 17, 2018 [30 favorites]

I was just gonna post the same thing as tarshish bound. In that article they take pains to note that they were the closest of the big models. You don't remember how much abuse they were taking at the time from people who wanted to believe the Princeton Election Consortium's 98% Hillary chance or the Times's ~90% Hillary chance?

I was one of the many who desperately wanted to believe the Comey letter wouldn't make an impact, but when I was on the ground for Hillary in Ohio, I got an inkling that it was a bigger deal than I had hoped. What happened was a highly correlated Midwest overperformance by Trump that it seems to me must have had a lot to do with that letter, since it is clear that late deciders broke for Trump, after all of the polls were out of the field.

You expect 538 to be able to account for decisions made after all their data was collected?
posted by lackutrol at 9:41 AM on August 17, 2018 [4 favorites]

Mods: Want to fix "528" in the post title?

I say leave it. It's within the margin of uncertainty.
posted by Frayed Knot at 9:44 AM on August 17, 2018 [90 favorites]

They have both Iowa 1 and Iowa 3 as lean D. Please, please, please, please, please.
posted by ArbitraryAndCapricious at 9:49 AM on August 17, 2018 [1 favorite]

We haven't had primaries up here in NH-1 yet, and I wish I knew how they decided that Maura Sullivan was going to be the presumptive Democratic nominee. I'm guessing it's entirely or mostly based on cash on hand?

We've got 11 declared candidates right now. Sullivan was basically hand picked for the district and moved here recently (like, last year) so while she's got a lot of national party support, she's real vulnerable to "You don't even go here!" attacks (see also Levi Sanders, who's getting this for living in NH-2). Chris Pappas has got the local party machine behind him, as far as I can tell, and is, at the very least, ahead on the lawn sign game.
posted by damayanti at 9:49 AM on August 17, 2018 [1 favorite]

> Right around the chance that 538 was giving Trump right up to the election, in fact.

> this is still not that encouraging a projection: it gives Republicans a 1 in 4 probability to keep control of the House. That’s not too unlikely.

Yeah, Nate Silver and Galen Druke emphasized this when presenting these results in their podcast. Three in four chances of regaining the House looks good for Democrats, but it's not remotely the same as in the bag. We shouldn't get complacent. (I can't believe I got through making this post and forgot to include that note of caution. It's something I meant to emphasize.)
posted by nangar at 9:52 AM on August 17, 2018 [1 favorite]

That "How the House has swung historically" chart: wow, did Gingrich have a big impact. Look how long the House was Dem before then, and how rarely since.
posted by doctornemo at 9:56 AM on August 17, 2018 [3 favorites]

How many votes does Russian Intellignece get to change on electronic ballots? Asking as a curious and appalled network security professional.
posted by Abehammerb Lincoln at 9:57 AM on August 17, 2018 [8 favorites]

(Thanks, mods, for fixing the typo.)
posted by nangar at 9:57 AM on August 17, 2018

Right now, the forecast has the GOP at 25.4% to retain.

For reference, Bryce Harper’s batting average is .243 this season. So the GOP has a little better chance than the chance he gets a hit his next time up. He has 101 hits this season.
posted by Huffy Puffy at 10:05 AM on August 17, 2018 [1 favorite]

Does 538 take into account its own influence on people who read 538 and then adjust their likelihood of voting? I mean, if they say 90% possibility of X candidate winning, do people therefore consider it a done deal and then stay home? Seems like every single one of their predictions should include a footer which says "...IF EVERYONE WHO READS THIS PREDICTION ACTUALLY VOTES"
posted by JohnFromGR at 10:06 AM on August 17, 2018

wow, did Gingrich have a big impact. Look how long the House was Dem before then, and how rarely since

Gerrymandering is a big factor too, particularly since 2000. Nationally the Democrats have to win by 7% or more of the popular vote to get 50% of the congressional seats. This was explicitly engineered by an open Republican effort.
posted by Nelson at 10:09 AM on August 17, 2018 [16 favorites]

Count me as another one to whom 538 is dead following the 2016 elections. The particular image I remember that sums up how misleading they were is their "winding path".

I don't know how a person can look at that image and not think that 538 was strongly suggesting Clinton would win. And maybe other outlets weren't any better, but to me that's not a convincing reason to follow 538; it just means that polling is worthless or even harmful to pay attention to.

Trump's election was traumatic for me, and for many people, in ways that we're still dealing with. 538's coverage contributed to the sense of shock that made his victory more traumatic than it had to be. I won't be fooled again.
posted by The Minotaur at 10:26 AM on August 17, 2018 [3 favorites]

After their catastrophic miscalculation with the 2016 election polling, Five Thirty-Eight just needs to go away and stop pretending they're relevant anymore. You drop the ball that badly, you go home.

A R G H

The thing to remember is that 538 doesn't say who will win, but gives you odds. The thing about odds is, unless they're 0% or 100%, they can't really be falsified. So long as your chances are at least 10%, 1%, 0.1%, any%, you have a chance of winning.

And Nate Silver is a statistician. He knows what odds mean, he uses them with precision. If Trump had a 20% chance of winning, that's a roll of 1-4 on a D20. Any D&D player will look at that and think: if it's a chance to miss an attack, I'll take it; if it's a saving throw vs. death, let's try to get out of it. Because rolls of 4 do happen, and critical fails are one time in 20.
posted by JHarris at 10:30 AM on August 17, 2018 [43 favorites]

Yeah, the responses in this thread are ridiculous. If you thought 20-25% meant "impossible" and were shocked when it turned out it meant "fairly unlikely", that's on you.
posted by howfar at 10:32 AM on August 17, 2018 [60 favorites]

That's why they play the game. None of this is decided on paper.
posted by whuppy at 10:40 AM on August 17, 2018

FWIW I disagree a bit that it's on the reader who misunderstands the 20-25% chance and thinks that's "impossible".

Instead we should blame
1) an education system that doesn't teach probability very well
2) a media environment that promoted the idea that 20-25% meant impossible
3) human brains for being bad at interpreting odds
4) a little bit 538 for insufficiently communicating that 20-25% is way too freaking likely when you're weighting a disaster, particularly as readers are used to reading polls where a race that's 75-25 *is* very unlikely to go the other way. The distinction between poll numbers/odds is important, 538 does make it, but readers who only get the topline number evidently don't.
posted by nat at 10:42 AM on August 17, 2018 [16 favorites]

I want to believe.

But I remember 2016.

Chance of my changing my mind: too close to call.
posted by Halloween Jack at 10:45 AM on August 17, 2018

November 4, 2016: “Trump Is Just A Normal Polling Error Behind Clinton”. Harry Enten, FiveThirtyEight

Popular vote, but still.
posted by Huffy Puffy at 10:47 AM on August 17, 2018 [6 favorites]

Also as someone who really wanted to believe the Princeton Election Consortium's number, I remember talking to some colleagues a week before the election, saying that even 1% chance was way too damn high.

I feel the same way about the House now. A 1% chance of effectively losing our democracy is completely unacceptable, a 25% chance is downright frightening, and now I'm gonna get back to work so I have time to write some postcards tonight.

In short: No numbers are good enough. Work instead of reading the numbers.
posted by nat at 10:47 AM on August 17, 2018 [13 favorites]

I wonder how many of the people heaping abuse on 538 now for underestimating the probability of a Trump win are the same people who were heaping abuse on 538 before the election for playing up his chances. The party line on Metafilter, as I recall, was that Nate Silver was intentionally and cynically portraying the race as closer than it was to drive traffic. Sam Wang of the Princeton Election Consortium, who was claiming "greater than 99%" odds for Clinton, was supposed to be the honest, straight-shooting alternative. Some short memories in this thread.
posted by enn at 10:57 AM on August 17, 2018 [59 favorites]

Polls are trailing indicators in the first place, and aggregating polls even more so, which is a definite problem dealing with an October surprise, yet 538 still outperformed your supposedly leading indicators like pundits and betting markets.

Still, with the House you've got a lot of races that are simply not going to get publicly polled at all, which is going to limit the quality of the predictions.
posted by ckape at 11:18 AM on August 17, 2018

I don't understand people who hear a statistician say "A is more likely than B" and then immediately want to tell the statistician how stupid they are when B happens.
posted by 256 at 11:25 AM on August 17, 2018 [25 favorites]

There are a lot of people who have heard the saying 'There are three kinds of lies: lies, damned lies, and statistics.' and instead of taking it as a clever line, take it as pure truth.
posted by Quonab at 11:29 AM on August 17, 2018 [2 favorites]

Guys, could we talk about 2018 and stop fighting about 2016? If 538 is dead to you, you could just...not comment. That’s a thing.
posted by schadenfrau at 11:30 AM on August 17, 2018 [30 favorites]

Nelson: If like me you lean Democrat and live in California, we have a lot of particularly exciting close races. Josh Harder in CA-10 Manteca, Katie Hill in CA-25 LA/Ventura, Gil Cisneros in CA-39 LA/Orange County, Katie Porter in CA-45 Orange County, and Harley Rouda in CA-48 Orange County are all running very close races. They could use your help! CA-04 Sierra, CA-07 Sacramento, CA-21 San Joaquin Valley and CA-49 Oceanside are also competitive. I'm also a supporter of Audrey Denney in CA-1 Chico / NE California; she's more of a long shot but she's great. I'd love to have her representing me in Grass Valley.

Democrat (well, socialist) in Cali here! These are all races I want to send my dollars and time to, because I want California to be as blue as possible from the Sierras to the Pacific. And I'm in CA-11 where our very well liked progressive representative has virtually a 100% chance of re-election, so no drama there. And Gavin Newsom and Dianne Feinstein have pretty much a lock on the Governor and Senate respectively. Helping elect more D's to the House is where my energy is best spent.
posted by Rosie M. Banks at 11:34 AM on August 17, 2018 [4 favorites]

Looking at the map, if we can flip WA-03, CA-45, and CA-49, then Democrats could hold every single inch of the Pacific shoreline of the Lower 48. We shall fight on the beaches!
posted by mbrubeck at 11:37 AM on August 17, 2018 [7 favorites]

(and flipping Alaska is also possible, in a wave election!)
posted by mbrubeck at 11:39 AM on August 17, 2018

Guys, could we talk about 2018 and stop fighting about 2016? If 538 is dead to you, you could just...not comment. That’s a thing.

Making sure people understand what the statistics meant in 2016 seems pretty darned relevant to trying to understand what they mean now.
posted by straight at 11:47 AM on August 17, 2018 [10 favorites]

Interested to see how PA fares this year now that we've been un-gerrymandered. My guy, Doyle, is running unopposed as usual but I'm watching the Lamb - Ruthfus race in what's now PA-17 and happy to see that Lamb has a 93% chance.
posted by octothorpe at 11:50 AM on August 17, 2018 [1 favorite]

I'm just not interested in any polling NOW for the election to be held in 87 days. "If the election were held today" has zero meaning when anything can happen in 3 months. Not to mention how a "big event" that could sway the electorate could take time to sink into the electorate's collective mind (especially when it's Officially Too Early to think about it). Come back to me a week before the election.
posted by oneswellfoop at 12:08 PM on August 17, 2018 [1 favorite]

One thing I found out in November 2016 is that there are a hell of a lot of people in my life who think 75% chance of winning means "will definitely get 75% of the vote." That kind of innumeracy is shocking to me, but my social media was filled with that. (Some from despairing Democrats, a lot from Trumpists shouting "suck it, libs!")
posted by Pater Aletheias at 12:21 PM on August 17, 2018 [16 favorites]

The thing that should be banned is that goddamn New York Times needle thing. Watching that around 10:00-11:00 election night was a memorable part of the trauma.
posted by lackutrol at 12:23 PM on August 17, 2018 [10 favorites]

Huffy Puffy: "November 4, 2016: “Trump Is Just A Normal Polling Error Behind Clinton”. Harry Enten, FiveThirtyEight

Popular vote, but still."

See also:

Sept. 1st: As The Race Tightens, Don’t Assume The Electoral College Will Save Clinton

Overall, Clinton’s leads in the tipping-point states — the ones most likely to determine the Electoral College winner in a close election — average about 4 percentage points, close to her numbers in national polls.

Oct. 31st: The Odds Of An Electoral College-Popular Vote Split Are Increasing

But Clinton is performing worse than Obama in 10 of the 12 states that were generally considered swing states in 2012. In some cases, such as Florida and Pennsylvania, the difference is negligible. She’s underperforming Obama substantially, however, in Iowa, Michigan, Ohio and Nevada and to a somewhat lesser extent in Wisconsin and Minnesota. She’s considerably outperforming Obama in Virginia and North Carolina, conversely, but that’s not enough to make up for her losses elsewhere.

Nov. 1st: Yes, Donald Trump Has A Path To Victory

If you want to debate a campaign’s geographic planning, Hillary Clinton spending time in Arizona is a much worse decision than Trump hanging out in Michigan or Wisconsin. Sure, she could win the state — but probably only if she’s having a strong night nationally. If the results are tight next Tuesday instead, Michigan and Wisconsin are much more likely to swing the election.

Nov. 3rd: Why Clinton’s Position Is Worse Than Obama’s

The point, as we’ve said before, is just that Clinton’s so-called firewall is not very robust. If you’re only ahead in exactly enough states to win the Electoral College, and you’d lose if any one of them gets away, that’s less of a firewall and more of a rusting, chain-link fence.

It's hard to argue FiveThirtyEight soft-pedaled Clinton's vulnerability given multiple articles like this, plus their small-but-hardly-impossible estimate of a ~33% chance of Trump victory. And that was *before* the Comey letter, which Silver believes was enough to tip the election all by itself.

If you want to grouse about misleading forecasts, focus on the Princeton Election Consortium, which gave Clinton a >99% chance and led to founder Sam Wang eating a bug on national TV.
posted by Rhaomi at 1:04 PM on August 17, 2018 [32 favorites]

If like me you lean Democrat and live in California, we have a lot of particularly exciting close races. Josh Harder in CA-10 Manteca, Katie Hill in CA-25 LA/Ventura, Gil Cisneros in CA-39 LA/Orange County, Katie Porter in CA-45 Orange County, and Harley Rouda in CA-48 Orange County are all running very close races. They could use your help! CA-04 Sierra, CA-07 Sacramento, CA-21 San Joaquin Valley and CA-49 Oceanside are also competitive. I'm also a supporter of Audrey Denney in CA-1 Chico / NE California; she's more of a long shot but she's great. I'd love to have her representing me in Grass Valley.

Emphasis mine bc I want to talk about CA-07 a little. It's my hometown (Folsom to be specific), and anecdotally (bearing in mind that I haven't lived there in 13 years) I would say it's purplish leaning conservative. Folsom more so than places like Elk Grove and Citrus Heights, but still. (That said, most of the alt-right types I know/knew from growing up around there were actually from CA-01, heh.) It's currently represented by a Democrat, but it was highlighted as a seat that Republicans think they can flip; that article is from March and things are looking much better now (538 has it as Solid D), but as people in this thread have noted, we can't assume anything. The current rep is Ami Bera, an Indian American physician who won the seat in 2012, beating out Dan Lungren, who famously opposed reparations for the Japanese American internment. All of this is to say, this race is very, very personal to me.
posted by sunset in snow country at 1:10 PM on August 17, 2018 [5 favorites]

Not to mention how a "big event" that could sway the electorate could take time to sink into the electorate's collective mind (especially when it's Officially Too Early to think about it).

Well, at least that's less likely. Presidential elections are about a person, and it's a lot easier to come up with something to be-FUD a single human being than a horde of them across the country.

What I really hate about elections now is that the Right has pretty well settled on their strategy over the past few decades, and it's entirely despicable. October surprises and cooking up controversies! They come up with the most asinine things to claim their opponent is Of The Devil. Like Benghazi. Like "Fast and Furious," remember that? They find anything they can so that, when the next Republican takes office and something horrendous happens, they can lessen the cognitive dissonance of all those voters by pointing and saying "It's them too!" And that tricks enough people into thinking both sides are the same that progressives can never ever get any actual progress going.

It's sickening, and it's going to happen again and again. When you don't care about truth and have people like James O'Keefe in your employ, you'll rig up anything to try to discredit the other side. Mark my words, things that seem silly now will somehow be trumped up (no pun intended) into Ultimate Cosmic Evil.
posted by JHarris at 1:16 PM on August 17, 2018 [1 favorite]

"If the election were held today" has zero meaning when anything can happen in 3 months.

While I understand your scepticism (which is warranted), it is worth noting that models like and including the ones used by 538 are predictions of what what will happen on election day, rather than what would happen in a notional election held today. Of course, the data they're working with are asking how people would vote today, so trying to use that to make a prediction for what will happen in 3 months time is a genuine challenge, but it is at least one they're trying to meet.
posted by howfar at 1:30 PM on August 17, 2018 [2 favorites]

I think we should remember that a Democratic wave, if it happens, will consist of Democrats voting, volunteering for and donating to Democratic candidates in big enough numbers to flip gerrymandered districts currently held by Republicans. What Nate Silver the other stats analysts at 538 are trying to do is predict the likelihood that we'll do that in big enough numbers to make that happen based on data they have like generic ballot polling, sporadic polling of individual districts and donations reported by individual candidates. So far, they're saying it looks there's a good chance we will, but it's far from certain.

For US Democrats, and Democratic leaning independents who want to see Democrats take control the House, state legislatures and governorships to prevent bad stuff from happening, this is not some impersonal force of nature we're talking about. We're going have to make the blue wave happen. Vote! Volunteer! Donate to Democratic candidates! Check your registration – because sometimes people disappear from the rolls! Apply for an absentee ballot if you need one! That's how we'll make a blue wave happen, or not. It all depends on what we do.

The main thing Democrats should take away from 538's analyses is where to concentrate resources in this election. Nathaniel Rakich has a list of 25 races based on their analysis that are critical for the Democrats to regain control of Congress. If you're a Democrat in a safe blue district and are wondering how you can help, you might consider donating to some these candidates.
posted by nangar at 1:50 PM on August 17, 2018 [10 favorites]

Silver on ABC News: “Democrats are not yet and may never be in a zone where anything is to be considered safe or taken for granted”
posted by Huffy Puffy at 2:51 PM on August 17, 2018 [3 favorites]

Still, with the House you've got a lot of races that are simply not going to get publicly polled at all, which is going to limit the quality of the predictions.

I'm just not interested in any polling NOW for the election to be held in 87 days. "If the election were held today" has zero meaning when anything can happen in 3 months.

Well, they've built in a lot more non-polling data than their presidential forecasts use, they've come up with a system for simulating polling data to plug holes in the actual polling, and they've also done away with the now-cast precisely because "if the election were held today" is unhelpful and potentially misleading. They've put a heck of a lot of thought into this, as can be seen in their methodology article which is the second link in this post.
posted by Anticipation Of A New Lover's Arrival, The at 2:59 PM on August 17, 2018 [2 favorites]

or even if there’s a relatively modest systematic polling error in the GOP’s favor on Election Day

It's not like crazy conspiranoiac to absolutely expect this at this point, right?
posted by rodlymight at 3:08 PM on August 17, 2018

It’s pretty nuts that the Democrats are forecasted to win the national House vote by almost 8 points and yet only have a 75% chance of winning the election.
posted by Automocar at 3:19 PM on August 17, 2018

Like if they win by that large a margin and the Republicans maintain control of the House, I’m not sure we can still say we live in a democracy.
posted by Automocar at 3:19 PM on August 17, 2018 [4 favorites]

Polls don't account for the impact of voter suppression, Russian hackers, and other ratfuckery. I am not making a cynical joke. This will happen.
posted by qxntpqbbbqxl at 3:24 PM on August 17, 2018 [4 favorites]

I was surprised to see how badly Rodney Davis (R-IL, downstate district including Springfield) is doing in their predictions (67% chance or so), and that's before polling prices in his latest clusterfuck, where one of his campaign employees gets drunk while being a tracker at his opponent's event, threatens the female candidate, and then starts throwing punches and screaming "THIS IS A HATE CRIME!" while being thrown out of the event. There's video. (Note that this is why Davis won't have any town halls, because he's afraid of people being "rowdy" and "disruptive.")

This seat is a real stretch for Dems, but the Londrigan (the dem candidate) is outraising Davis locally (Davis has more PAC money from national GOP groups), Davis refuses to commit to any debates, Davis refuses to do any town halls, and now his campaign keeps own-goaling. So it's heartening to see the 538 forecast is that bad for Davis, and maybe can get worse ... especially as he keeps providing video for ads.
posted by Eyebrows McGee at 3:24 PM on August 17, 2018 [5 favorites]

It's really hard to believe polls, or even have faith in the electoral system, when an 11 year old kid can hack into an election website in a matter of minutes.
posted by basalganglia at 3:28 PM on August 17, 2018 [1 favorite]

Making sure people understand what the statistics meant in 2016 seems pretty darned relevant to trying to understand what they mean now.

Also, there were many faux-sophisticated voters who said 2016 was a safe election to vote third-party, skip it, write-in Bernie, etc. even in close states. Even if many of them switched to saying the polls were wrong after the fact, that attitude needs to be actively debunked until there’s more universal civic education or some sort of electoral reform, both of which will only happen if people go out and vote for Democrats even if they think the conclusion is foregone.
posted by adamsc at 3:48 PM on August 17, 2018 [5 favorites]

nangar: I think we should remember that a Democratic wave, if it happens, will consist of Democrats voting, volunteering for and donating to Democratic candidates in big enough numbers to flip gerrymandered districts currently held by Republicans.

Exactly this. We've seen what happens when Democrats vote in their numbers. The "Blue Wave" is US, the voters, not something that some benevolent force like the DCCC or whoever will do on our behalf. The worst thing we can do is cower in fear of the Unstoppable!Republican!Juggernaut! and throw up our hands in resignation, or, worse, stay home.

I think a lot of us are still finding our feet after years of "if we have the Presidency that's what really matters" and ignoring down-ballot races. But now that we are awake and voting, look at what we're accomplishing - the all-women D tickets in Minnesota and Michigan, Conor Lamb's win in Pennsylvania, Democrats running in seats they never really considered before.

And something really personal for me, in my own district, for State Assembly - my Lamorinda neighbors keep sending the execrable, and Republican, Catharine Baker to the state Assembly. Since this is a sky-blue Democratic district, I don't know why. Her D opponent is Rebecca Bauer-Kahan who I badly, badly want to see elected, because even a "moderate" Republican is still a Republican and a blue district doesn't need R representation.
posted by Rosie M. Banks at 3:54 PM on August 17, 2018 [1 favorite]

Obfuscation about what he's actually saying is Nate's bread and butter, though; since elections are unrepeatable events, it's meaningless to assign them a probability at all, much less days/weeks/months in advance. He's talking about how often one candidate beats another in a computer model he's developed; a reliable relationship between that model and actual election results is something people take mostly as an article of faith. It certainly hasn't been established with anything like mathematical rigor.

In short we have and can never have any idea whether Nate was "right" in 2016 when he said Trump had a 25% chance of winning; the whole thing is hokum top to bottom.
posted by gerryblog at 4:08 PM on August 17, 2018 [1 favorite]

But since he's predicted more than one race in his career, you can look at overall past success rate. The last article quoted has a list of predictions for all 435 House races -- if it's hokum, their hit rate should turn out to be no better than random, which I very much doubt it will.
posted by hoist with his own pet aardvark at 4:15 PM on August 17, 2018 [4 favorites]

Well, he's been at it for ten years already; where is that analysis? Are we judging him on whether or not he picked the candidate who won to have a greater than 51% chance of winning, or are we judging him on whether or not 1 in 4 of his 25%-chance candidates win? And are we judging him based on his "last prediction," or on a running average of every prediction he made in advance of each election?

How do we know which of the 100+ <50% predictions of Trump possibly winning was "right" when we say "Nate got it right in 2016"? We don't even know if it's really true that Trump was just bad luck, or if he actually wins in every parallel universe that runs that election...

He's using real information so I doubt he'd be wrong more than he is right, but the reputation he's developed as a guru is a clever branding strategy predicated on nearly all readers of his site not understanding the nature of the claims he's making and not holding him accountable for whether he's right or wrong.
posted by gerryblog at 4:28 PM on August 17, 2018

How about calling every state correctly in 2012? That strikes me as pretty impressive. I don't know how his hit rate compares with e.g. Sabato's, but it shouldn't be that hard to quantify.
posted by hoist with his own pet aardvark at 4:35 PM on August 17, 2018 [1 favorite]

calling every state correctly in 2012

If he says x% chance of winning, though, he should be "wrong" 100-x% of the time over the long haul. 100 hits and zero misses suggests the odds component doesn't actually map onto the election in any meaningful way; if he's making a meaningful prediction about the odds of winning, then we should actually see the less likely candidates win pretty commonly, roughly in accordance with how often he predicted they would.
posted by gerryblog at 4:41 PM on August 17, 2018

And (just to add to that) you can see how the motivated reasoning of the 538 fans works here: when the candidate he said had a greater than 50% chance of winning wins, Nate was right because he called it; when the other candidate wins, Nate was right because he only ever assigned it a probability. There's nothing falsifiable about what Nate is doing because it's all about (very carefully) crafting a brand.
posted by gerryblog at 4:44 PM on August 17, 2018

since elections are unrepeatable events, it's meaningless to assign them a probability at all

this is an extremely controversial doctrine! faithful Bayesians would disagree strongly.
posted by vogon_poet at 4:45 PM on August 17, 2018 [5 favorites]

gerryblog, it really sounds like you have a problem with the field of statistics in general rather than with Nate Silver in particular.
posted by Anticipation Of A New Lover's Arrival, The at 5:02 PM on August 17, 2018 [22 favorites]

Once again:

538 popular vote prediction: 49/45
Actual popular vote: 48/46

And elections are most definitely not independent events. As one example, the winner last time can run again as the incumbent, which grants an advantage. extrrnal factors like the economy also both repeat, and play a roll.

I don’t know what beef you’ve got with 538, gerryblog, but you’ve grossly mischaractetized all of the arguments made in this thread. “Nate said greater than 50% so he got it right” is in fact the exact opposite of what’s being said.
posted by Frayed Knot at 5:04 PM on August 17, 2018 [10 favorites]

Mod note: Let's return to talking about the material at the links and consider the objections to 538 and so on fully aired.
posted by Eyebrows McGee (staff) at 5:16 PM on August 17, 2018 [4 favorites]

faithful Bayesians

What's in a Church of Bayes? I mean, altar cloths embroidered with Bayes' Theorem can be assigned a high-value prior probability, but what's the liturgy like?
posted by tobascodagama at 5:49 PM on August 17, 2018 [2 favorites]

Well, it depends.
posted by Huffy Puffy at 6:26 PM on August 17, 2018 [13 favorites]

Presbyterian, presumably.
posted by lefty lucky cat at 6:38 PM on August 17, 2018 [1 favorite]

(and flipping Alaska is also possible, in a wave election!)

I sure hope so! Don Young has been representing Alaska since 1973. (This is longer than I’ve been alive, and ... I’m not very young any more, alas.)

I am super excited about Alyse Galvin. I’ve met her several times; she’s very thoughtful, and she will do a good job representing Alaska.
posted by leahwrenn at 7:52 PM on August 17, 2018

Silver's book is all about how people want oracles, not statisticians, and don't understand, or refuse to understand, how statistical predictions work. He's the first to admit that "calling" all 50 states correctly in 2012 was a statistical fluke, as good at it was for his career, and there are confounding factors for this model that make it much more difficult to predict in a way that people would prefer. I wouldn't treat it as a source of truth - and neither, I think, would Nate Silver want you to treat it that way.

The value here is in looking at what factors 538 consider to be of interest for this race (we don't need a statistician to tell us obvious things like getting more people out to vote is a viable strategy and is probably not priced into the polls properly). Things like how Rohrabacher is considered a likely 'tipping-point district', that is, if you ordered the districts by their margins, Rohrbacher's district is more likely than most to be smack bang in the middle. It's a district that's marginally more important, because it's more likely to swing - it's probably also a district where there's going to be hijinks. Or how Republican incumbents don't seem to benefiting much from their incumbency basically anywhere, which suggests the 'run everywhere' strategy has statistical support.
posted by Merus at 9:06 PM on August 17, 2018 [10 favorites]

What I really like about the way this is done, and about 538 in general, is the transparency. I've not encountered another news organization that is so willing to pull back the curtain and explain their methodology, the assumptions they're making, and the reasons why they made the various judgement calls that they had to make. It's almost as if they think their readers are intelligent adults who can be trusted to handle ambiguity and uncertainty, rather than children who must be told to trust the authorities. It's like they actually have confidence that their work will hold up under scrutiny, like they know they've been diligent and careful and can stand behind their work.

What other news organization can you say that about? Basically none, in my experience. Part of the reason why people don't trust the news media—and remember please that trust in news media had been declining for years before Trump came along to pour fuel on the fire—is that lack of transparency, that appeal to authority, and the fact that when we do learn about journalists' methods, they don't always seem diligent and honest. Journalism as a profession (in the US anyway, can't really speak to other countries) needs to step up its game, do careful, thorough work that it can stand behind, and start showing its work. I imagine that most news organizations would find that absolutely terrifying, and rightly so. But if you're terrified of having someone see how you do your job, there's a good chance it's because you're not doing it very well.
posted by Anticipation Of A New Lover's Arrival, The at 3:52 AM on August 18, 2018 [5 favorites]

On a different note, I'll be really interested to see how things shift after Labor Day, when more polls will start using likely voters rather than just registered voters. We may start to see that enthusiasm gap appear in the polling, if there is one.
posted by Anticipation Of A New Lover's Arrival, The at 3:55 AM on August 18, 2018 [1 favorite]

Democrats would have a problem if there’s a shift in the national climate toward Republicans

cue october surprises campaign
posted by flabdablet at 7:09 AM on August 18, 2018

I think the correct plural is octobers surprise.
posted by C'est la D.C. at 7:21 AM on August 18, 2018 [3 favorites]

I have a certain amount of confidence that this administration will attempt some sort of surprise but will manage to fuck it up badly and end up looking worse.
posted by octothorpe at 7:41 AM on August 18, 2018 [2 favorites]

I take some solace in knowing it's probably too late to start a major war ahead of Nov 2018. Perhaps he's holding that move in reserve for 2020.

Back to 538, the one thing missing from the site is detail pages broken down by state and/or district. As fun as it is looking at the whole US, House elections are still very much a local affair. 700,000 people elect each congressperson and each race is different and local.
posted by Nelson at 8:28 AM on August 18, 2018

There won't be an October Surprise, just like there wasn't one in 2016. (Comey's letter was released in November.)

Look out for Sweeps Week instead.
posted by tobascodagama at 10:44 AM on August 18, 2018

One day, when this era is all in the history books and fully automated luxury gay space communism has finally asserted itself, I shall distribute gifts to deserving parties on randomly selected days in October; by living thus shall I take my revenge.
posted by howfar at 3:09 PM on August 18, 2018 [2 favorites]

Comey sent his letter to Congress on October 28; Jason Chaffetz tweeted about it the same day.
posted by hoist with his own pet aardvark at 6:15 PM on August 18, 2018 [2 favorites]

Hello! Today is the last day to register to vote as a Democrat in NYS if you are currently unregistered. Register here.. The primary is Thursday September 13th. NO IDC lists primary chellangers to the republican backed IDC “turncoat” Dems, check to see if they’re in your district. Cynthia Nixon/Jumanee Williams are endorsed by the DSA as well as Julia Salazar in Brooklyn/Queens. Zeypher Teachout would make a spectacular DA and has made going after Trump and corporations and monopolies a priority.

Primaries are won by very few votes, so yur vote counts a lot this September.
posted by The Whelk at 8:11 AM on August 19, 2018 [3 favorites]

Also “...O’Rourke’s energy is palpable, infectious; his sweat is the physical evidence of that energy leaving his body. And it seems to be working. Even as he struggles with a continued lack of name recognition, in a state that has consistently voted Republican for the past three decades, recent polling places O’Rourke just two to six points behind Cruz. Among volunteers, there’s cautious yet barely contained glee: Could O’Rourke pull off an upset that, just six months before, seemed impossible?” Anne Helen Petersen follows Beto O’Rouke around Texas ahead of the midterms
posted by The Whelk at 8:44 AM on August 19, 2018 [4 favorites]

I'm loving watching Beto work in Texas, and donated to his campaign, but he's still a bit of a long shot. For Senate races, Kyrsten Sinema in Arizona and Jacky Rosen in Nevada are also worth following or supporting. Both women are looking to flip Republican Senate seats and both races are pretty much literally 50/50.

Unfortunately 538 doesn't have a page (yet?) on Senate races, but here's some alternative prediction sites: 270 to win, Real Clear Politics, Cook Report. The bad luck in 2018 is that of the 35 Senate seats up for election, 26 are already held by Democrats. So there's not a lot of room to gain ground. Most predictions I see say the Republicans are guaranteed 48 or 49 seats and the Democrats 44 or 45. That leaves 6-8 seats up for grabs.
posted by Nelson at 9:14 AM on August 19, 2018

Obligatory reminder that while she will be much better than McCain or Flake, you will not like Sinema very much and should not expect to. If she wins, when she shows up she will be the most conservative Democrat by a decent margin and you should not expect anything else.
posted by GCU Sweet and Full of Grace at 9:40 AM on August 19, 2018 [3 favorites]

I'm with Liam Donovan's thoughts on the Texas Senate race. O'Rourke is quite unlikely to actually win but the enthusiasm behind his campaign could have a significant impact on downballot House races. I'd been concerned that we were kind of wasting all those resources on tilting at windmills but it's money well spent if it builds up the Democratic infrastructure in Texas and nets a couple House seats, even if O'Rourke ends up 5 points short in the end.

Re: 538. They nailed 2016. Absolutely nailed it. To the point that, as I've said previously, I changed Nate Silver's name in my bookmarks to Nate Gold. Their analysis of 2016 was "if the polls are accurate Clinton will win. But even a standard polling error in Trump's direction means he would win and here's how..." and then 538 went on to show how a slight overperformance among white working class voters in the Rust Belt would give Trump a plausible path to an Electoral College victory.

Aaaaand then Trump slightly overperformed among white working class voters in the Rust Belt and followed exactly that path to an Electoral College victory.

538's key insight to the 2016 was that polling errors would be correlated. The other analysts were saying that even if polling errors in one or even two of the states Clinton needed were wrong, she had leads in enough states that it wouldn't matter. But Silver&company realized that if polls were off in one direction in one state, they were far far more likely to be off in the same direction in other similar states. So if polls missed in Michigan, as they did, it meant they were far more likely to miss in the same direction in Wisconsin, and Ohio, and Pennsyltucky.

Now people are like "well duhhhh" but the truth is usually obvious in hindsight. It's recognizing it in advance that is hard. And they did, and nobody else did.
posted by Justinian at 11:22 AM on August 19, 2018 [21 favorites]

What I do think 538 could do a better job of is the thing that Sam Wang (who, to his credit, still has the final 2016 presidential election result at the top of his homepage, as a monument to his failure) tries hard at, which is informing voters, activists and donors where their efforts are best spent. Predicting the outcome really doesn't matter all that much to campaigns or activists; although it has implications for lots of people whose interests depend on aligning themselves with government policy, I don't think those are the people driving the vast majority of 538's traffic. What is important to the interests of the majority of people checking 538 ten times a day is what they can do, although I do recognise that they may not be able to or want to do anything beyond spectate. And while it's not 538's job to promote activism per se, I think the site could become more useful for the people who use it if it pushed the analysis of individual races into the foreground of reporting.
posted by howfar at 12:21 PM on August 19, 2018