Nate Silver puts 538's reputation on the line with last minute hedging and a shrug.

[Fiendish Thingy]

Here’s Nate’s long-winded apologia from yesterday where he essentially says they’ve been manipulated by the flood of recent shitty R polls, but since they’re in the prediction business, they’re going to predict any way:

https://fivethirtyeight.com/features/the-3-big-questions-i-still-have-about-election-day/

And here’s today’s article, where he continues to hedge, basically saying “anything could happen, but whatever happens, we were right”:

https://fivethirtyeight.com/features/final-2022-election-forecast/

I know he’s obligated to ABC to provide clear, numerical predictions to drive the dumbed down horse race narrative, but with all the anomalies this year, not just in polling, but in so many other factors, it would have been more honest to use 538’s reputation to say an accurate prediction was impossible to make this cycle.

[iemanja]

1. You'll find out in a few hours

How accurate the polls are. I don't see the point of complaining about them now.

[Fiendish Thingy]

2. I'm not complaining about specific polls

I’m complaining about how 538, with its huge audience, allowed itself to be manipulated into distorting the horse race narrative into one of an inevitable GOP victory.

Nate chose clicks over integrity.

[iemanja]

3. No, you're complaining about all the polls

In an effort to convince yourself they are wrong.

[Fiendish Thingy]

4. You seem to be ignoring the fact that the GOP deliberately manipulated polling averages

And 538 continued to make predictions based on those manipulated averages.

Thats what I’m complaining about.

[iemanja]

9. I don't buy it

but the more I see people make that claim, they more they show me how nervous they are.

[Fiendish Thingy]

17. By "they", I'm assuming you mean the GOP

Yes, while everyone was looking for a single incident “October surprise”, the GOP was providing a string of surprises with a flood of polls from Trafalgar, Rasmussen, Insider Advantage and other Republican sponsored polling outfits that provided outlier (Walker beating Warnock by +6? Seriously?) results compared to the very few polls by reputable nonpartisan polling outlets released in the last couple of weeks of October.

And 538 and the majority of the news media fell for it, gobbled up 538’s corrupted averages, and unquestioningly regurgitated the GOP’s desired narrative.

That is some Roger Stone on steroids level ratfucking right there.

[iemanja]

11. Most polls are well within the margin of error

Which means their results aren't statistically significant.

[DLCWIdem]

15. Yes

Even after he found out about all those GOP leaning polls hre said his model took that I to consideration. But there was a lot of GOP leaning polls. His model takes a couple of right leaning polls into consideration but there were a lot of those polls so how could his model compensate for all those polls.

[MiniMe]

6. I think it will be more like a few days

[Wibly]

12. Don't think so

Poster is not complaining about the polls. Poster is pointing out the polls have been manipulated and that this particular organization, 538, has been impacted and is now trying to cover its butt.
Silver and his crew have been unable to provide any sort of data on new registrants and their leanings, and have relied, pretty much wholly, on the suspect polls being put out by organizations linked to partisan groups. 538 has also leaned heavily on previously registered, and party committed voters, and have no data on what is going on with the previously uncommitted and unregistered.
I also think, based wholly on your reactions here, that you are have a knee jerk reaction to being told that the data you are leaning on may not be accurate.
Just thought I'd throw my two cents worth in here. No offense meant.

[Hassin Bin Sober]

19. That's how I've felt for the last week or more.

We are about to mostly find out.

[inthewind21]

21. No we won't

It will be days. What we will know tonight is how many people voted in person ON election day and early in person. We won't know anything about mail in.

[budkin]

5. I stopped following him after 2018

He's more of a pundit now than anything.

[LakeArenal]

7. Yes.

[gab13by13]

14. I agree

[CousinIT]

8. I have ignored 538 for several years now...

...they got 2016 SO SO wrong. I'm done with them. And with MAGAts manipulating the pools of polls so they can scream "FRAUD!" - nobody knows WTF will happen.

[Fiendish Thingy]

18. 538 was actually the MOST accurate in 2016

They gave Trump 1-in-3 odds of winning, more than any other pollster or media outlet, IIRC.

[Beautiful Disaster]

10. I get why. Trafalgar Group is 100% a Republican pollster.

But there is no doubting their likely voter model led to some pretty close results in 2016 and 2020. They were pretty accurate. If you're Nate, are you wondering if these partisan polls are just manipulating and getting lucky or actually on to something with their likely voter models?

Will be interesting to see.

[SlimJimmy]

13. Likely voter models are always more accurate, no matter who does them.

I've been screaming this here for months, but I guess people just want to have the result they want.

"... or actually on to something with their likely voter models?"

[Fiendish Thingy]

20. Except this cycle, LV models could be way, way off.

First of all polling response rates have dropped below 1% (was 30% in early 2000’s); in addition many new registrants and first time voters get excluded from LV screens, and this cycle appears to have a larger than usual share of first time voters,

[SlimJimmy]

30. Half of registered voters don't vote.

And most of the major pollsters were using registered voters even up until last week.

[Beautiful Disaster]

26. Plenty of LV models were off in 2020.

Specifically in states like PA, WI and MI.

Take Wisconsin. LV models from different outlets were woefully off.

Like Emerson, a fairly respected pollster, their LV model had Biden winning 53-48 - an eight-point margin. The New York Times; final LV poll had Biden up 11!

Trafalgar had Biden winning by one.

So, even in 2020, there was a huge gap between LV models and Trafalgar was far closer than established polling organizations like Emerson, Siena and Marquette, which had Biden by 5.

[SlimJimmy]

28. Trafalgar has a method of finding voters that don't respond to other pollsters.

They call it the neighbor model. Instead of asking who "you" are going to vote for, they ask who do you think your neighbor is going to vote for and are they likely to vote. It sounds simple, but it seems to work. And for the record, they use likely voters in nearly all their polling.

[Beautiful Disaster]

29. I am not saying they don't - I am saying other LV models have been way off.

As I showed you.

So, just because it's a LV doesn't mean it's fully accurate.

[SlimJimmy]

31. I'm not saying fully accurate, I'm saying more accurate than using registered voters as the model.

[Beautiful Disaster]

32. It's a pointless point, tho, because no one is talking RV models.

So, I'm not sure why you're making it about LV/RV. In fact, RV models are rarely used this late in the campaign. They only do RV models early on when it's harder to determine LV models. Every poll that's released nowadays is likely, if not certainly, a LV model.

[SlimJimmy]

35. RV polls were used by polling companies as late as last week.

I'll quote some if you like. All I have to do is go back to my posts from a week ago where I was ranting about it.

[Deminpenn]

25. A broken clock is right twice a day

That's how you should view Trafalger. Their model assumes aggrieved white men are undersampled in polls, so they weight to favor them. Since this group is a GOP base demographic their model favors Rs. In 2016 and 2020 they were right because Trump drew out these exact voters in large numbers. But this is something completely unique to Trump.

[Beautiful Disaster]

27. It's very possible you're right.

But so far, the clock hasn't proven it's broken just because their model heavily leans into a more Republican electorate.

The closest we have is 2018 and Trafalgar was a bit more hit & miss but certainly not wrong across the board.

Take Michigan in 2018. They had Stabenow winning by 9 - she won by 7. Not a huge difference.

Conversely, they had Cruz winning Texas by 9 and he won by 3 (rounding up).

Trafalgar had Kemp winning Georgia in 2018 by 12 points - he won it by 1.4.

Conversely, Trafalgar was the only final poll from Florida that had DeSantis winning over Gillum - and he won.

Trafalgar did have Laxalt winning the Nevada's gubernatorial election - and he lost. But also had Tester winning in Montana.

So, 2018 was definitely not nearly as neat for them as 2016 and 2020. But they weren't horrifically off like their polls appear to be now.

I think this is the election that makes or breaks these Republican pollsters. If they whiff, and the Democrats win Pennsylvania, Georgia, Nevada, Arizona and maybe even Wisconsin, I think there's going to be a lot of questions about counting these polls in the future.

But we'll see.

[ThunderRoad]

33. Trafalgar credibility question?

Well researched. So, what's the call on Trafalgar if they have a good showing tonight? They certainly haven't been afraid to publish outlier results across many high profile elections this year. And they've been consistent going way back this Summer. If they prove right again tonight do they go from a pollster we ridicule to one we take seriously?

[W_HAMILTON]

34. Trafalgar has gotten 66% of their races right -- that's terrible.

[gab13by13]

16. Here is what I have to say about the polls

they extremely underestimated Democratic enthusiasm.

All of you poll backers explain to me how every poll had the Kansas abortion question rated as a dead heat, every fucking poll.

What were the results? The abortion issue won by 18 points. Every poll was off by 18 points. Couple that with all of the special elections that were also either dead heats or the Magat favored, Dems won those special elections also.

To put in perspective; Fuck the Polls, they are used for propaganda.

[getagrip_already]

22. he makes his living off polls - what else is he going to do?

He is stuck with a very good buggy whip in the 1st century.

He did have one very good tool though back in 2016. It allowed you to change the demographic makeup of voter turnout to see the effect it could have on the election.

Moving black voters up, even by a lot, had little impact on the national results (they turn out at such high numbers there is little marginal opportunity). Same with other minority groups.

What had a huge impact? Increasing the number of white voters. Even a tiny increase had a large relative impact.

Guess what happened.

That tool appears to be retired. It's probably more useful to pollsters doing what-if's than it is to us, except it was an eye opener and explained tfg's white nationalist turn.

But nate is just trying to survive. His website has had yuuge sponsorship problems. Guess he is just chasing the dream.

[BannonsLiver]

23. 2008 seems like such a long time ago.

[nolabear]

24. He's so yesterday.