Survey USA and TargetSmart are junk

[molova]

Hillary is not losing by 7% on North Carolina and not winning by 8% in Florida.
1) SurveyUSA is a robocall
2) TargetSmart is a huge outlier. Florida is close

[BluegrassDem]

1. Did you know watch O'Donnell tonight?

Targetsmart actually polled those who already voted

[writes3000]

2. Did you watch the TargetSmart pollster on Lawrence O'Donnell?

I wouldn't be so quick to dismiss their findings. They might be off but there's some specific, credible work they're doing.

[stopbush]

3. No, Florida is not close.

[molova]

4. Yes it is

It is.

[MyNameIsKhan]

7. No it is better than both polls u described above

I can give to non stop points why it is:

Republicans with over enthusiasm are voting early and eating their Election Day turnout ..,

Hispanics are covering for AA votes

We only need 13% of AA votes to win with current trend compare to 16.5 Obama won in2012.

Let me know u need more

[Demsrule86]

37. Polls show 28% of early GOP vote went to Clinton...it was a poll

of people who already voted and were asked how they voted...

[stopbush]

8. Nope. You will be joining the media pundits on Nov 9

who have totally misread what is going on in this year's election, which is going to be a landslide victory for Hillary that will include her carrying Florida by a wide margin.

[molova]

20. Feel free to call me out on my ignorance on November 9th

Please remind me of how silly I was for not predicting a +8 margin in Florida.
Let's just hope the egg doesn't fall on your face.

[Demsrule86]

38. The fact you have no interest in a very important poll about Florida comprised of

people who already voted is surprising or not.

[duffyduff]

46. And you believe Trump will win

That is what you are saying.

[molova]

56. You just lied

I never predicted such a thing.
Think before you post.

[Wounded Bear]

49. He didn't predict a +8 win...

the +8 number was based on exit polling of current early voters. He actually said it will probably drop of a few points by election day, but would still end up with FL going Clinton by +3 to +5.

I can live with that. Would love it if it were enough to oust Rubio, but I'll take what I can get.

[molova]

57. Link? Nt

[Wounded Bear]

60. No link...

but Lawrence O'Donnell's show might be on the MSNBC website.

[molova]

58. By the way, you misrepresented the methodology

The TargetSmart poll is not based solely on exit polling of early voters. It's also based on whom they consider likely voters.

The combination of Hillary's 17% really vote lead and those likely to vote on the future yields a result of a final 8% lead for Hillary.

[Wounded Bear]

61. I didn't say "solely"....

but since my post was rather terse, I suppose it could be misinterpreted to mean that.

[Foggyhill]

5. This comment is pretty weak...

They're polling actual voters, thats the absolute opposite of junk
She won't win by 8 but it's not 1% either
So not close
She's been polling around 3% for weeks using old likely voters models
The Latino vote is strong which means more than 3% with gotv

[MyNameIsKhan]

6. I do statistics for profession, targetsmart is on mark

This is the best you can get ... Even myself I cannot fathom +8 HRC but can understand with early voting data ... We are for sure in +2 category ... Rest +6 of the margin, I need more data. Which they promised to post tomorrow. Btw +2 in Florida is 160k votes ... I rely this poll more than +6 Siena poll and +6 SurveyUsa poll today ... They are junk ...

[Foggyhill]

10. The reason it's good is because your closing

In on actually randomly sampling the voters (since you don't have Election Day voters, it's still a bit off for the final score but it's getting there)

The analysis part relies on the sampling part being ok. Usually it's not, the methodologies are usually very flawed

[Hortensis]

13. Thanks for the discussion.

[Fahrenthold451]

54. Was looking at the internals.

They claim more accurate than an exit poll. I agree based on the wealth of data that Florida provides. That early vote data combined with the uncast ballot cohort = a pretty solid number.

[Divine Discontent]

9. I don't dispute that FL is no cakewalk. Here are the numbers though that you should include after

your comment.

TargetSmart/William & Mary
Florida Presidential Poll - 530 Online interviews & 188 phone interviews ---- 48-40, Clinton +8


But, as for their Early Voting Poll - 311 people were polled after voting ---- 55-38, Clinton +17

I also would like to know where they did this early voting poll. Obviously, the calls and online interviews presumably are from all over the state. However, I would think the early voting was done in a general area, but hopefully it was from all across the state. I just have reservations that it was done that way, but I put some faith in those exit poll numbers they obtained from 311 people, but feel it would have been more reliable with a few hundred more, and it would have been around 10 points, well below the 17 they claimed their exit poll showed, even so, if she's up by 10 now, and she loses election day's votes by 1%, then she still wins by nearly 9%.

[Wounded Bear]

50. The full report is supposed to be out today (Monday)...nt

[Farmgirl1961]

51. I would be good to have more information on how they did their study

While I read that Nate Cohen rated TargetSmart as a C rated poll, I also read over at Daily Kos that the kind of analysis they're using is new/cutting edge (so maybe it can't be properly rated yet).

I'm hoping that once the full study is put out that folks here who are a lot more knowledgeable about this kind of stuff and pour through it and give the rest of us some meaningful analysis.

[Divine Discontent]

53. thanks!

[bigdarryl]

12. Why are people like you are.so dome and gloom?

[Cicada]

14. Single polls often capture noise so we should use average of all polls

With so many polls there will be some far off the mark due to chance alone. Sometimes a coin comes up heads 5 times in a row.

[Adrahil]

17. Quite true.... but....

This one is interesting in that it specifically targets people who already voted as part of its methodology. Do I think HRC is really up by 8? I doubt it, but this is good news for us, no matter what.

[Cicada]

18. Great methodology but very small sample size of only 331

28 percent- plus or minus 8 just for one standard deviation. Encouraging but not very reliable.

[Adrahil]

19. I agree, to a certain extent.

I don't think HRC is up by 8. And it's unlikely she is really winning 28% of Republicans. But the fact that happens at all is NOT a good sign for Cheeto.

[BainsBane]

15. Their sample size is much larger than most polls

I don't believe Target Smart is junk at all.

[molova]

21. Sample size does not determine accuracy

No expert ever said that.

[MyNameIsKhan]

27. So NYT Siena poll interviewed about similar number

Of republicans and that is a better poll right?, the method is almost identical ... Check voter rolls to create likely voter screen and weights.

In a poll containing 900 samples only 350 would be republicans...

Do you know what was sample size in entire exit polls for Florida in 2012? Few thousand people.

[Demsrule86]

39. The smaller the sample size, the larger the margin of error...

[Persondem]

42. Sample size IS the variable that most determines accuracy.

The formula is ...

Z = (stand. deviation)/ n^(1/2)


n is the sample size, and z* is the appropriate z*-value for your desired level of confidence

[duffyduff]

47. And YOU are an expert?

Tell us your qualifications.

[titaniumsalute]

16. Another senseless, baseless post

No thought, no reason, no logic, no debate. Just an empty, stupid post.

[molova]

22. Tell us ,titaniumsalue, what your forecast is for Florida

And remember, don't say anything far from 8%.

[titaniumsalute]

23. Mostly sunny, about 82 degrees

[molova]

24. I referred to the forecast for the US election results in Florida

That's the thing with people who defend outliers. When confronted, they get shy about forecasting the results yielded the outlier.
It's almost as you really don't believe TargetSmart.

You're better off talking about the weather.

[titaniumsalute]

28. You post something without any real thought or point

You are entitled to your opinions...just as I'm entitled to ask you why you might believe in a certain way.

The problem is EVERY poll is an outlier at this point because we have no idea what the actual benchmark is. You cannot really call something an outlier without knowing a benchmark. The individual state polling has been so sporadic, volatile, and statistically bouncing that we don't know at this point is an outlier is an outlier. We WILL know on Nov 9th.

I've done survey measurement for years. A poll is a poll. The same company can take the same poll of a different sample with changes to the results. It is getting more difficult to poll people as there are some many different ways in which you need to reach them. Landlines, cell phones, online, etc.

[molova]

30. To say that a poll is a huge outlier requires no further elaboration

Stop pretending that I have to give you an op-ed about how nobody else is close to showing an 8% race in Florida.
You want Hillary to win by 8%. So do I. Heck, I wish she won by 40%.

But you're pretending TargetSmart is not an outlier.

[PunksMom]

26. Hahaha! Nice!

[molova]

31. Punks, do you believe Hillary will win by 8%?

What will the margin be in your opinion?

So far the pattern is this: People defend TargetSmart but don't predict an 8% victory themselves.

[duffyduff]

48. Do you believe GOP voters are all going to vote party this time?

This is not a typical presidential election.

[Demsrule86]

40. And one wonders why it is so important to this poster

to knock down a poll that is actually a poll of those who have already voted.

[workinclasszero]

25. Your concern is noted

[molova]

29. Your irrational belief in outliers is duly noted

There.
Now watch: I'm going to ask you whether you think Hillary is going to win by 8% or so, and you're not going to answer.

Do you believe Hillary will win Florida by 8%?

[MyNameIsKhan]

33. No that would be difficult ... That amounts 800k votes

I think 200k votes is along the line ...

But then no one questioned Bloomberg/Selzer +6 in Ohio that meant suddenly HRC will not only loose Obama margin but 6 points down that is about 400k votes, so she even ends up under Kerry numbers in 2004. There is hardly population change in Ohio past 4 years.

Nate happily added in his data with good weight.

So every poll is showing state of race .., target smart poll shows HRC having comfortable win.

[Demsrule86]

35. Your irrational belief that every good poll for Democrats is an outlier is duly noted.

This is a poll of those who have already voted...again...this sort of post make many here suspicious of the motivations of the such posters. And if these results hold...as voting continues...yes Hillary Clinton could win Florida by 8% give or take the margin of error.

[molova]

59. You lied about me doubting "every good poll" for Democrats

Drop the Straw man.

[duffyduff]

45. There is going to be a LOT of crossover voting. This election is an anomaly.

Two major factors are at work: a totally unfit and unqualified candidate is running on the Republican ticket, and the possibility of the first woman president in American history.

To ignore the obvious differences between this campaign and previous ones means you have no understanding of what is going on.

Few Dems will cross over or vote third party. MILLIONS of Republicans, especially the women, will cross over to the Dems, leave the top of the ballot blank, and a few might vote third party (especially in Utah).

[still_one]

32. Unlike other polls, this polled people who already voted. Over three million

people in Florida have already voted.

What it found was of the registered republicans who voted early, 28% crossed over and voted for Hillary. With in that same 28%, they also voted for Rubio, not Murphy. They also noted a significant increase in early voting among Latinos. Those factors alone pushed the odds toward Hillary.

They also projected that if that pattern follows it will favor Hillary.

Hillary, the campaign, or most of DU is NOT taking Florida or any state for granted.

That is why a lot of us are doing phone banking into the swing states, Florida, North Carolina, etc.

For the OP to characterize the poll as junk, is not based on any factual data. In fact for the early votes its data is accurate than most, because it actually polls the people who voted, not just looks at registration numbers and assumes, republicans voted for republicans, or Democrats voted for Democrats.

Whether this pattern will apply to the general election in Florida, that can be debated, and why a lot of us are working are asses off calll banking

[Demsrule86]

34. I disagree about Florida...I saw the pollser on O'Donnell last night

He has polled those who already voted...and 28% of GOP types are voting for Hillary...is it the Cuban vote...we don't know, but we do know that the Hispanic vote is up...The election will be over by 9:30 if this is true. And the polls that show a 'closing' are using intensity...just as the Romney era polls did...and they were wrong...there is plenty of intensity both for Hillary Clinton and against Donald Trump within the Democratic vote.

[geo1]

43. Here's the data

Maybe already posted, but just in case

[link:https://www.scribd.com/document/329698329/TargetSmart-William-Mary-Florida-Poll-of-Early-and-Likely-Voters|

[Avalux]

44. Well, thank you for setting us straight. Are you a statistician?

Provide evidence to back up your assertions; in particular TargetSmart. You haven't done that at all in this thread. What's your deal?