Selzer pollster predicted Romney would beat Santorum by 9% in Iowa (2012)

[Molusko]

Santorum won by 0.1%.

That's for those who liken Selzer with God. They not always nail it. The praise for this pollster is nauseating.

http://www.realclearpolitics.com/epolls/2012/president/ia/iowa_republican_presidential_primary-1588.html

[elleng]

1. Thanks for the info.

[elleng]

2. I just learned

'She noticed an upward trend--even a surge--for Santorum during the polling period. I'm paraphrasing here, but she noticed his numbers increasing at a high rate during the polling period and she did predict a surge for Santorum.

She released her numbers with that footnote and she noted that Santorum was surging.'

[CoffeeCat]

3. Are you ready to be more nauseated

Wow, people are really trying to discredit Ann Selzer and her Iowa poll this evening. What's up with that?

This was explained on other threads, but Selzer released the data from the Republican 2012 race that included Santorum and Romney.
That's correct that she said that her data showed Romney beating Santorum by 9.

However, she released the data with a caveat. She noted that her data showed a Rick Santorum surge that increased with each polling day. Thus, she declared that Santorum was surging and she formally declared the race a toss up.

So--while it's true to say that she released those figures (with Romney beating Santorum by 9) that is not the entire story. To omit her "caveat" is to distort her record. Because, in the end she called it, spot on.

She said Santorum was surging and the race was a toss up. Santorum ending up winning the Iowa caucuses by less than 50 votes.

Why are some on this board trying to discredit Ann Selzer's Iowa Poll the night before she releases her poll?

It's all very interesting.

[BainsBane]

4. No one poll is itself terribly meaningful

The averages tend to be most informative. Regardless, we'll know soon enough. Monday is only three days away.

[CoffeeCat]

5. The Iowa Poll is meaningful

From FiveThirtyEight--1/27/16
Ann Selzer is the Best Pollster in Politics-- how her old-school rigor makes her uncannily accurate
http://fivethirtyeight.com/features/selzer/

The Iowa Poll really is "meaningful" to use your word. Her knowledge of Iowa and the caucuses as well as her science-based methodology--make this the most meaningful poll that is released in the caucus season. There's a reason that she holds a press conference to announce her numbers. These numbers are important.

She's crazy accurate and she predicted Obama's 2008 win, among others.

[BainsBane]

6. Fine

I know nothing about her. All legitimate polls use scientific methodology, though some have better track records than others. My point is the same regardless of the pollster. Averages are more informative than any single poll.

Evidently you think the poll will tell you what you want to hear. Regardless, we will know soon enough what the actual results are.

[CoffeeCat]

7. Actually I know nothing

about what The Iowa Poll results will be. I am not anticipating any particular outcome. Just accuracy.

And also, not all polls use scientific methodology, as you said. There were so many polls this season that were horrendously flawed. They were junk polls. That's the problem. That's why I hang my hat on this one, because Selzer is so hell bent on accuracy.

Averages, at this point--are not helpful when it comes to Iowa. So much is volatile, due to the caucuses happening so soon. Polls that were taken in Iowa two weeks ago--throw them out. Too much has changed.

That's why this Iowa Poll is important. It's a snapshot. But it should give us so much information. And for all I know, it will show that Bernie is behind by 10. I have absolutely no idea.

You said that you know nothing about Selzer. This article came out two days ago from fivethirtyeight.com "Ann Selzer is the best Pollster in Politics; how her old-school rigor makes her uncannily accurate"
http://fivethirtyeight.com/features/selzer/

[winter is coming]

8. IMO, they were deliberate junk, meant to skew the average. n/t

[CoffeeCat]

9. Well, when someone like myself can read about their methodology and who they polled

and spot obvious flaws--it's pretty pathetic.

At worst, it's rigged polling meant to intentionally deceive the American people. At best, it's a bunch of lazy, unprofessional pollsters who should find another line of work.

In the end--it doesn't really matter what their motivations or intentions were. The polls were junk. So any outfit that uses aggregates with those junk polls in them--is no longer reliable or valid.

[BainsBane]

10. The volatility is precisely why averages matter

And even a first-rate pollster can have an occasional skewed sample. Silver has frequently talked about the importance of averages over individual polls. Silver doesn't incorporate junk polls into his statistical models.

Caucuses are particularly tricky because turnout is more difficult to predict. The polls that show Sanders doing the best--like the CNN one last week--depend on high turnout from people who have never caucused before. It certainly could happen, but it does make modeling more challenging.

[Betty Karlson]

12. Trendlines may tell more than poll results by themselves. eom