If The Polls Are So Bad, How Have We Forecasted 101 out of the last 107 Senate Races Accurately?

[tgards79]

Some facts:

Polls are not perfect -- how could they be? - but they are very accurate. Dismiss polls at your peril -- they tell you where to focus your volunteer efforts.

As for the barrage of GOP polls -- the aggregators and forecasters recognize this phenomenon, understand it and adjust for it. And it's not that those polls are necessarily wrong -- they were generally more accurate in 2020 than the mainstream ones -- there are just a lot of them. That is easy to adjust for....Nate Silver devoted part of a podcast to this topic if you want to know how.

The key to reading these "dire" forecasts is to get motivated by them and ACT. This election is going to be very, very close. Democratic volunteers should focus TODAY on GOTV in the Senate races in GA, NV, AZ, NH and PA. We MUST keep the Senate as it is very likely we are going to lose the House.

Stop whining about the polls and GET OUT THE VOTE!

For a great up-to-the-minute forecast of every single Senate, House and Governor race, read this:
http://www.borntorunthenumbers.com/2022/11/btrtn-2022-official-midterm-election.html

[betsuni]

1. K&R

[Amishman]

2. Because polling and people's behavior has changed

Go read Wasserman and others, pollsters can't get responses and voting patterns are so disrupted their turnout models are barely better than guesses.

People used to answer the phone. Now, thanks to the endless onslaught of scammer phone calls, no one answers. Scams through text or email have us paranoid and annoyed about any unsolicited contact.

That and there are relatively few competitive races each year. We have four tossups and two close leans this cycle - that's somewhat unusual

[tgards79]

3. Well then...

You think none of this was true in 2016, 2018 and 2020? Of course it all was...and we still got 101 out of 107 races correct. People have not been answering phones for years. Pollsters supplement phone with text and online surveys.

For the record, there are six toss-up Senate races: GA, PA, NV, AZ, NC and OH and both NH and WI are on the edge between toss-up and Lean. That is EIGHT very, very close races.

[Amishman]

4. unfortunately OH and NC are far from being tossups

I'm sorry for the harsh words, but this is reality.

Budd is up by an average of over 5% in the last 5 polls in NC. That's not a tossup.

Ohio is worse, The past 8 polls are all over 5%, with four of them being 8+. That is definitely not a tossup.

From someone who is arguing that the polls aren't so bad, I am surprised to see this disconnect.

We can still hold the senate, but a large pickup is probably off the table. Best to understand this now. Everyone still needs to go vote, as systematic polling errors can happen - but even then they are unlikely to be 5+% errors in our favor across multiple states which is what we now would need for a big night.

Why is polling especially harder now? There is no good comparable data. Polling requires a turnout model to match responses against the expected electorate. With the pandemic, Trumpism, voter suppression, and normalization of mail-in voting - the data from 2016, 2018, and 2020 is less useful. How and who voted in those elections is of limited use in projecting because each year has seen major changes in the mechanics of the election.

[Sympthsical]

8. People like to note the polls being off in 16, 18, and 20

However, they like to leave off the second half of that sentence - the polls typically way over-counted us and under-counted Republicans.

Are they over-adjusting now? Did they not adjust enough? We'll find out today. But in recent history, we've been showing up in polling a lot more than Republicans have been. We talk about people not answering polls. Well, if I look at the two sides, which one do I think is far more suspicious of the media and doesn't want to talk to them? I don't think it's us. That's why paid surveys are always heavily tilted Democratic. We happily give our opinions to whomever. The Right is weirder about it.

I think polls are in the ballpark, but in races so tight, ballpark can still be an either way proposition.

Control of the Senate could go either way (I'm slightly optimistic there, but I agree the House is lost), but I expect most results to be within their error margins. If we're over-represented for the fourth election in a row, pollsters are really going to have to start asking themselves questions. They can take a mulligan on 2020, but 2022 has been fairly typical.

[Claustrum]

18. +1

There is a part of me that understand why the media is tilting on the side that republicans will win this year (based on historical trend that party in power loses and the polling errors in recent elections). There are signs that we could lose both. But as noted by a lot of DUers and Nate Silver himself that polling errors could go either way, we really don't know how things will shake out this year. I've not followed much any polls (definitely not looking at them on a daily basis like I did in 16, 18 and 20) this year.

But anyways, I still think the narrative that the media used is not correct. For a party in power while the economy isn't great, we should be looking at like losing a few seats in the senate and 40+ seats in the house. The fact that this is anywhere remotely a close race should spell trouble for the republicans.

[Sympthsical]

19. That's why I think Dobbs is already in the polls

In a typical midterm year with this inflation, Republicans would be comfortably ahead in polling, even with the under-counting we've seen in recent years.

That they're not indicates to me that an unusual push back has been detected by the numbers.

Some people think there's a stealth wave a la Kansas, but Kansas didn't have that much polling. This election, however, has been polled to death. If there was any great stealth factor, I think it would've been picked up. I think it already has. That's why these close races and close generalized national sentiment fairly evenly split when it ordinarily should not be.

We'll see. I'm very open to being wrong about it.

[Claustrum]

20. For Kansas, I think the fact that it's a standalone issue helped too.

If it was a race between a democrat who supports abortion rights and a republican who supports banning abortion, I think the result would be extremely different. We've seen this in polling for years. People might be more progressive on issues when they are standalone but then they shift to supporting republicans when it's attached to a candidate.

[Sympthsical]

22. That's true

How many other issues - if you just put that one issue as the election - would turn out much differently than if it involved a Republican vs. Democrat race?

"Today we shall vote only on minimum wage"

It wouldn't even be close.

[The Revolution]

25. Florida 2020

Voted 60-40 to raise minimum wage to $15.

Support Democratic/Liberal positions, but vote for Republicans. It is a strange phenomenon in some ways.

[Sympthsical]

26. But what if the politicians were removed

Because I think a lot of issues are a matter of prioritization.

Isolate an issue, and you'll find some people who really, really, really care about it. Then there are people who are really opposed. Then there is a third category who have a firm opinion, but it's not so important that they'll be specifically motivated to weigh in on it.

I have all kinds of opinions on things, but I don't share most of them unless they're specifically solicited. Now tell me I have to leave my house to go vote for one thing. Will I leave my house for goat cheese? I don't like goat cheese. Do I care enough to leave the house to vote against it? Not really. Doesn't affect me or I don't care. However, if I'm already there voting on soda vs. juice (which I have an invested preference for), I will probably also tick a box registering my opposition to goat cheese. My opinion is being solicited while I'm worried about other things.

D v. R puts in different motivation incentives for showing up where other votes can be solicited for things people may care a lot less about. I will never leave the house to vote against goat cheese. I would, however, vote today, tomorrow, and next week if there was a vote strictly on Free Coffee For All.

If that makes sense. I think I got lost in the example somewhere.

[Demsrule86]

24. Time will tell...and I don't care what you say. I don't think the polls are accurate. We could still

lose but it doesn't matter in terms of polls....trashing this thread.

[W_HAMILTON]

28. When you take into account the margin of error of these polls...

... even a five point lead would be considered a toss-up since almost every poll has at least a two or three point margin of error (which you then apply to EACH candidate's polling total).

[iemanja]

30. Is 5 points more than twice the margin of error?

If not, it's a toss up.

[Amishman]

35. depending on the poll, but roughly yes.

And Ohio is a bit more than 5.

[iemanja]

36. That must be a very good poll

because most are 2.5-3.7 MOE. There are few polls with MOE's less than 2.5.

[GreenWave]

6. Oh the times they are a changing

GOP has a never ending supply of dirty tricks. Push polls. leaning right to get averaged in. Under 1% response.

Media not asking Big Questions: Why would anybody vote for a GOP rep who tried to overturn Biden's victory on Jan 6th? Why was that not the national narrative?

And you did no go back deeper in time either. Empty your mind and relax. Go back deeper in time. Stop at LBJ's landslide. Utter decimation of a nuke loving opponent. And how do you make money off your polls then? How do non-existent 24/7 political panic news make money?

This is why they want 50/50 and the polls are tightening up comments. landslide equals going out of business.

[iemanja]

31. You mean they aren't asking questions relevant to people like you

Why would people "even vote for the GOP" is not a valid question to anyone but those who despise Republicans. Seriously.

There is no point arguing about polls now. We'll find out the results soon enough.

[GreenWave]

32. But Jan. 7th, 8th the MSM did focus on GOP defectors.

Maybe I conflated two topics. If so my apologies.

[iemanja]

33. I fail to see how that's relevant to polling

You want the media to read the polls and then say, "how can they even vote for the GOP after Jan 6"? What purpose would that serve other than to make Democrats feel better?

[Genki Hikari]

37. And people keep telling you

That pollsters don't rely on cold-call responses, but you keep acting like that's the only way they contact anyone.

It's not. They know that text messaging and voice mail can reach people who screen calls, and they use it. That gets responses when the other method won't. Maybe not a vast quantity of responses, but enough to get in the ballpark.

Really.

Tired of correcting this flat-out wrong nonsense.

[dsc]

5. How many of those races were close

I agree that the polls are better than people are giving them credit for, but if say 15 of those races were close (within say 2 points) then the polls would be 9 and 6 not exactly Nostradamus territory.

[tgards79]

7. Actually...

In close races (less than 5 points) we are 15 out of 21. Success not perfection. Polls are not perfect!!! How/why can they be? The whole purpose of doing this is NOT to predict outcomes (which is just for fun) but to focus volunteer/donor efforts. Without polls you'd be flying blind. How would you know the PA Governor race is NOT close but the PA Senator race is? It tells you where to spend a dollar or an hour. At this point, focus on the SENATE. The House is lost. If we lose the Senate because we tried in vain today GOTV for House races, we are idiots.

[Better Days Ahoy]

10. Read the comments again and again

You're not absorbing the points, especially about people increasingly not wanting to talk to pollsters.
No amount of uppercase and exclamation points is going to make your point.
Polls have not kept up with the electorate, demographics, emotional triggers or public sentiment about being polled. Numbers folks typically aren't exactly leaders in emotional intelligence and historical nuances.
Numbers also can be sliced and diced any which way. I was in corporate finance for 30+ years and know how to make the numbers dance to nearly any tune I was asked to play.
Your industry is an exercise in creative, self-actualized justification that is fast becoming out of touch and largely unnecessary.

Happy Election Day, where exit polls and actual tabulated results are the story, not flawed predictive analytics.

[tgards79]

15. On the other hand

Being a corporate finance whiz does not make you a statistician. Take it from me, a longtime corporate exec, corporate finance requires addition, subtraction, multiplication and division, and a healthy understanding of returns (IRR, DCF), which is not exactly rocket science, especially if you have models and underlings.

The fact that you are admitting to lying about your earnings is something the SEC should look into. Shame on you for cutting the numbers the way your bosses asked you to instead of standing up to them and reporting them as a whistleblower. The corporate finance folks I worked with as an exec and long-time board member would never have stood for this.

I can go all night with you on nuance and emotion. You are making an argument not backed in facts. Polls are not perfect, but in 2020 we got 33 out of 35 Senate races right (actually 35 out of 37, because we called all four Georgia races, the regular and the runoff accurately). We got 48 out of 50 states right (and 5 out of 6 districts) and all 11 governor races. THAT WAS IN 2020, the strangest year of election dynamics ever. YOU are the one who has not kept up, because people stopped answering landlines long ago.

Polling is challenged but to say it is "way off" is simply fake news.

[Better Days Ahoy]

23. Never said I was a "whiz", junior.

Also never said polling was "way off".
I said it was unnecessary, out of touch and a self-masturbatory exercise. Your comments support it.

And no shit, Sherlock: pollsters call both mobile phones and landlines. Technology teaching opportunity backfired, Sparky.

Go after the other posters, little dude. Lots and lots are reacting negatively -- with supporting facts -- to your three (3) identical initial posts. WTF three?

Try to reply to them without uppercase, hyperbole, misstatements (fabulous attributes for your blog integrity, BTW) -- and without clutching your little pollster pearls. I'll wait.

[FBaggins]

13. 15 out of 21 is terrible

You would get almost that many by flipping a coin.

[tgards79]

16. Last time I checked...

Coins had only two sides. The odds are 50/50. This is 71/29. That's is much better, but obviously numbers are not your strong suit!

[FBaggins]

17. "Slightly better than a coin flip" is lousy for political prognostication

Particularly when you're still including "gimmie" races by looking at any within +/- five points.

Completely rubbishes the notion that the polls must be reliable.

[panader0]

29. Who is the "we" you are referring to? Do you work for 538?

[iemanja]

34. That's the purpose of campaign and party polls

but not public polls.

[Genki Hikari]

38. Maybe you need to understand Margin of Error

And why it's always listed in a poll's data.

It's because the +/- means that there's a range of probability of X happening. If a candidate polls at 52% in a poll with a +/- 3% MOE, then that candidate is likely to get anywhere from 49%-55% of the vote.

IT IS NEVER 52% and that's that.

That's what too many don't get.

People, especially our traitor media, are also relying too often on polls that don't meet the +/- 3% gold standard. If it's over that, then the poll is pretty much useless. Ignore polls that don't have an MOE at or under +/- 3%.

It will save everyone a great deal of grief if they'd learn some mere basics of polling and stats.

[dsc]

40. I teach stats for a living, so yeah I get MOe

my point is, that it is misleading (to be kind) to say one has forecasted 101 out of 107 Senate races when 80 of them could have been forecasted years before they occured due to the partisan lean of the state. Predicting a Democrat will win in Hawaii or a Republican in Idaho isn't exactly difficult.

[karynnj]

9. The majority of races are not in question.

For instance, there is no reason to poll VT this year.

This year's toss ups are just GA, PA, and NV. There are about 5 others where there is a small chance of the underdog winning. So, as to getting a call right, assume you get 50 percent of the first three right and, say, 80 percent of the somewhat possible upsets right. That means you would miss only 2.5.

[Better Days Ahoy]

11. There you go -- thank you

Most polls are like consultants:
Expensive and unnecessary.

Hello from a former NJ res.

[Genki Hikari]

39. BS, and dangerously stupid BS at that

Smart candidates use pollsters to fine-tune their campaigns, such as what issues they need to start covering, places where they need to have some campaign stops to bolster the locals, and so on.

Only very stupid people would try to guess by the seat of their pants what they need to do with those things, rather than relying on people who study where numbers of people actually stand on things.

[Better Days Ahoy]

42. Thanks for playing -- too bad your erudition got buried

In a tertiary post.
Why not start your own post if you're this passionate, dear one?

[FBaggins]

12. We haven't

“Our” track record is lousy.

[SWBTATTReg]

14. And there are some that suspect that the GOP is misleading people by these numerous 'polls'

that they are pushing out the door. Trying to suppress the blue vote I guess. Doesn't matter to me, I don't look at polls (the only one that matters is the actual voting today (and of course all of the early voting, and mail in votes prior to this day).

[BlueInPhilly]

21. The fundamentals of polling have changed

I have said many times that polling is a dying industry. For better or worse.

Technology is changing incredibly fast, along with people's behavior. A discipline that heavily relies on unbiased response by a normally distributed population will go the way of Blockbuster.

It might take a few cycles until people realize that polls have become a crap shoot.

[Genki Hikari]

41. Funny, they've worked fine for the most part

And they especially work for companies doing all kinds of marketing research about what products to bring to market, or where to send more product, regionally or by time of year, and the like. Only people who have no clue about stats, which is all that polling and marketing are, could say they don't work, or that the fundamentals have changed.

I guarantee that the reason why X car dealership will receive 5 green gomobiles while the one across town won't carry any green models is thanks to market demographics (stats) telling them that green gomobilies are likely to sell at the first location, compared to the latter.

The reason one supermarket in an upscale neighborhood will have different products, and quantities of products than another supermarket just like it in another upscale neighborhood is because the stats have told them that.

And so on.

The stats haven't changed in the least. Cultivating the samples may have changed, but the math hasn't changed for a very long time.

That's why the remark about fundamentals changing is outright silly.

[BlueInPhilly]

43. I build statistical models N/T

[BlueInPhilly]

44. Not sounding silly now, am I?

The math would never change. The theories still hold. But the practice, as it has been in place for so long, is untenable. Maybe I should write a paper about it, then you can read my peer reviewed “outright silliness”.

[LetMyPeopleVote]

27. Forget the poills, go vote

Link to tweet