I am genuinely curious. Why do we put so much stock in polls when we ALL

[Ferrets are Cool]

know that we elect a president by electoral process? Unless it is specifically a poll in a swing state, how much does it really matter?
Thoughts?

[TreasonousBastard]

1. Seems to me most people don't. But polls do show trends, for whatever trends are worth.

[Laelth]

2. I would like to know your own answer to that question before I respond.

Certainly, you have a theory. What do you think?



-Laelth

[Ferrets are Cool]

7. My answer is that it seems somewhat useless to sweat over each and every % point of a

national poll. GOTV is the best weapon we have whether we are ahead OR behind. I just don't want to see complacency because of a national poll showing the Democrats ahead by 10-15 points. Especially with the extremely LONG lines we are going to see on Nov 3rd.

[Laelth]

12. That doesn't answer your question.

Why is it that some of us are desperately curious about each and every new poll that comes out? Either we’re just stupid (and that appears to be your answer), or you lack empathy and can’t imagine why some people do what they do.

Your choice.



-Laelth

[Ferrets are Cool]

16. Thanks for such nice thoughts to my sincere question

[Laelth]

24. Try a little empathy.

Imagine that the people that you encounter in life (especially on DU) are descent, reasonable people who have very good reasons for doing what they do. Then try to imagine what those reasons may be (and that’s precisely what I asked you to do above when I asked you to answer your own question).

I, personally, am suspicious of any post that gives off the “Why are people so stupid? Why can’t they be just like me?” vibe. That’s what I got from your OP.

We’re still on the same side. I value what you add to DU. Thank you.

-Laelth

[Ferrets are Cool]

34. If that is what you read from my post, I really feel sorry for you.

I have always enjoyed your posts. This exchange is over.

[Claustrum]

3. Polls tell a trend and what direction the race is going.

You don't look at that individually and separately from the state polls. If national polls are ahead by double digits, there is no way that the state polls don't shift toward Biden as well.

[badseedboy]

4. I am a bit shocked by my poll watching,

since I swore I wouldn't trust another one after 2016 and its poll failures.

[RKP5637]

9. Agree!!! n/t

[brush]

19. Did the 2016 polls really fail? Or did the repugs cheat in 3 states and steal the election?

And of course there was Comey's letter, after which the polls reflected an immediate shift away from Hillary. She still won though except for the repug cheating in those three states.

So my contention really is that polls are pretty accurate. Now the battleground state polls have Joe ahead in most them, and even in Florida of all places.

[Bettie]

5. Me? Because I want some hope

so I look to the polls to give me some. It shows trends, though, to be honest, who actually answers their phone anymore?

But it gives us a "temperature" of the nation as a whole.

[MaryMagdaline]

17. This

[Demsrule86]

6. Polls matter and they look very good for Biden. Trump has lost support and consider he only

won by a bit.

[Hoyt]

8. I'd rather our candidate is ahead in the polls, whatever limits they have.

I read somewhere that if a Democratic candidate wins nationally by 5% to 6%, it is highly likely they'll win electoral college. It's less likely, when the national margin is smaller, but wouldn't call it unlikely.

[infullview]

10. They will provide a measure of how much cheeting there will be by the the repugs

[cayugafalls]

11. Polls are only as good as the demographic polled and the truthfullness of those polled.

They can be an indicator, but depending on the execution, could be off, which is why they always have margins of error.

If done right, they can show a trend, but should never be taken as gospel.

Personally, I don't look at polls for more than a minute and I don't dwell on what one poll says vs another. I take it for what it is, an indication of those who were polled and nothing more.

Since I can't see who was polled and know nothing of their honesty, I put little time in contemplating their power.

Liken it to standing inside and having someone tell you it is windy outside. You will never know how windy it is unless you stand outside for yourself, but the verbal description can be an indicator that there is wind.

[abqtommy]

15. I totally agree. An ethical poll will yield usable information. A corrupt poll won't.

[Ferrets are Cool]

13. Thanks for all the thoughtful replies

It helps me to understand.

[N_E_1 for Tennis]

14. I look ...always a little skeptical...

People tell you what you want to hear sometimes.

[Cosmo Blues]

18. I thought we all learned in 2016

It's the state and not the national polls that matters, yes it takes a little more effort, but the surprising depressing exasperating night in 2016 at least taught me to look state by state. And the national poll wasn't wrong it was within the margin of error by the time November 3rd rolled around

[Silent3]

33. That's not the exact lesson to learn from 2016

Yes, state polls matter more than national polls, but they are less frequently done, so not a very good snapshot of fast-moving trends.

And it's not like state polls are totally, entirely disconnected from national polls. They roughly go up and down together. You're not going to have Biden rocket up to 90% in California and simultaneously drop to 10% in Alabama.

The bigger the national difference, the lower the statistical odds of Electoral College madness overriding the popular vote. You have to go back to a weird four-way race in the early 1800s to find a case where the EC disagreed with the popular vote when the popular vote margin was as large as 4%. The few other times the EC diverged from the popular vote was with a 3% or less margin.

If Biden is even up by 6% nationally, never mind the double-digit margins we're starting to see, it is EXTREMELY unlikely that EC goes the other way. (Not counting, of course, byzantine chicanery to steal the election, wherein nothing about the popular vote much matters anyway.)

[Cosmo Blues]

36. 2000 and 2016

Chicanery is happening. I can see something happening that boosts the liberal vote in California and the same thing makes Alabama turn against us, not that Alabama matters, and since their Democratic senator is not running against a child molestor I don't see him holding his seat, so we need to pick up five other Senators somewhere. I have also heard that we could win by 6 million and still not win the presidency, Republican legislators are threatening to pick alternate electors, I'll keep going state by state

[Silent3]

39. For the level of chicaney you're talking about...

...even state polls don't matter then.

"we could win by 6 million and still not win the presidency" is consistent with what I've already said about the size of the national popular vote.

[HelpImSurrounded]

20. For the same reason I read Harry Potter and the Lord of the Rings

It gives me a chance to fantasize about good winning over evil.

[Johnny2X2X]

21. Polls matter

It's all we have to go by. The state polls matter most though.

Bottom line is that there is no way Trump wins the EV if he loses the popular vote by 6+.

[edhopper]

22. I am looking aty the polls

in the swing States. The general polls matter in how it shows up.
If the big lead is all California and New York, it doesn't matter as much as if it is spread throughout the States.
Remember Hillary only lost the EC by 70,000 votes with a 3 million vote lead.
Trump will find an 8 million vote deficit hard to over come.

[C_U_L8R]

23. Because data.

[brooklynite]

25. Answer: they show trends over time...

Yes, a national poll doesn't give you a precision estimate about individual States. But when the national polls show consistent upward movement over time, that's data supporting Biden's ability to win the critical States he needs.

[moondust]

26. I question the methodology.

I should probably search for some answers but...if they don't poll cell phones then how accurate could they ever be? And unsolicited calls to cell phones may involve some airtime cost to the voter but is that even legal?

[LakeArenal]

27. I think there are ways to manipulate polls.

But low information types like sound bites and easy to understand results.

[Fiendish Thingy]

28. If polls didn't provide useful information, $millions wouldn't be spent on them

Of course some polls, like push polls, are used to shape opinions, but, by and large, polling is a long accepted, evidence based research method used to measure support for candidates, issues, products. Just as with any other research method, the methodology of a particular polls can be flawed and provide inaccurate results, but that doesn’t negate the method of polling all together.

British Columbia is having a provincial election right now, and it’s frustrating because they never poll individual races, only party preference, and the results are often way off because of numerous close races - they may still be right on voters’ party preference, but significantly wrong in how many seats that party wins.

US polling, with the additional focus on states and congressional districts, is much more successful at predicting outcomes.

[SMC22307]

29. "We" don't. I ignore almost every poll-related OP.

[hlthe2b]

30. Right now there are battleground state polls with increasing frequency--THEY are worth watching

National polls are (perhaps) useful for subgroup trend analysis (e.g., seeing which groups that voted for Trump that are seemingly turning back to Dems for example). But, yeah, the state-specific polls of battleground states are what matter.

[honest.abe]

31. If Biden wins the popular vote by 10%, he will win the election by a landslide.

That's why.

[Yavin4]

32. Below is a table of the national margin translated into the odds of winning the electoral college

If Biden wins the national vote by more than 5 basis points, then he has a more than 98% chance of winning the electoral college. Right now, Biden is ahead by a polling average of 9%.

If Trump wins the electoral college while losing the popular vote by 9%, then there was mass cheating or interference.

Link to tweet

?lang=en

[Ferrets are Cool]

35. Thanks again for your thoughtful and insightful answers

[LexVegas]

37. You tell us why you put so much stock in them.

[Ferrets are Cool]

40. If it is NOT clear that I do not, then I need to learn how to make my posts clearer.

[Statistical]

38. Well three reasons

1) National polls tell you the national mood. That is important for determining how red or blue a state is relative to the national average. The relative performance of a state compared to the national average is very useful to determine is a state is slipping or gaining. where to devote resources, when to pull out, etc.

2) While there is a slight conservative bias to the electoral college it is only by 2% maybe 3%. That means the tipping point state (the state that puts a candidate over 270) is usually

[Ferrets are Cool]

41. Thank you for your thoughtful answer

[Amishman]

42. Several reasons

There are more states in play this time around, making the overall trend more relevant than usual.

It's a good bell weather of momentum. National polls tightened before state polls in 2016. Final RCP poll average was 3.2 in 2016, and results in popular vote was 2.1. just weeks before it was more than double that.

If the devil is in the driver's seat and we get a 2016 repeat, we would the national polls to tighten soon. Every day joe's lead stays near it's current lofty 9+% level is less of a chance of that swing.

The last part is seeing if Trump can buck the challenger's advantage. Typically the challenger picks up a greater share of late undecideds. 2016 had more undecideds than usual, 2020 has less. A late tightening would a red flag that this election is not following normal trends at all.

[Happy Hoosier]

43. Data has to be analyzed in context.

yes, of course the actual states are what matters, but we have TONS of data that establishes that state and national polling data are strongly correlated. That doesn't mean one to one, but the data show that individual states generally have a correlation range. So, for example, If a Biden is leading in National polls by 10 points, that correlates to a specific range of individual state outcomes based on historical data. And if we get actual state polls, we can be constantly adjusting those correlations.

I mean... this is what data wonks do. But this is STATISTICS, not a crystal ball, so it implies a range of likely outputs with a particular probability distribution and confidence levels.

In other words, if Biden is leading by 10 nationally, he is very likely to win enough state EV's to win the election.