How much do you trust the polls?

[kentuck]

With all the talking heads talking about how much each candidate is up in the polls in the battleground states, does your blood pressure go up every time the polls change?

Do you grow more skeptical every time the polls change?

Or do you tend to believe that the pollsters are mostly correct in their prognostications? After all, it is scientific.

How much should we listen to the polls?

[RKP5637]

1. Zero! n/t

[whathehell]

3. I tend to trust them, but when they differ by more than

3 or four points, I begin to wonder.

[randr]

4. First voted in 68

Don't recall polls being wrong until W. That was also the first election where the exit polls got shut down.

[Drunken Irishman]

8. Exit polls have been wrong before...

They just typically happened in elections where one state didn't decide the whole thing like in 2004.

In 1976, exit polls had Jimmy Carter winning Illinois, and one of the networks actually called Illinois for Carter, even though, when the counting was done, Ford won it. It didn't cost Ford the election, though, as Carter won without Illinois.

Exit polls, like any poll, has a MOE and can be off. The exit polls in 2004 were way off, though, and didn't make a lick of sense. I believe the PA exit poll had Kerry winning by almost 20 points - which would've been a bigger margin than Obama in 2008 and Clinton in 96 - two elections where the Dem won in a landslide.

Kerry wasn't going to win PA by that margin. He just wasn't.

They were off then. I don't know how they got 'em wrong. But they were off. But it wasn't the first election to happen.

[underpants]

5. Well they make great sausages but their submarine and pistol designs are lacking

[kentuck]

7. I saw that on a sign...

someone had posted.

[Drunken Irishman]

6. Never trust just one poll...

Look at the collection of polls to get an idea of where the race stands. Polls are generally accurate in an overall average.

[LonePirate]

9. The 2016 MI Dem primary polls proved to me that polls cannot be trusted this year.

I think we may be on the verge of becoming an unpollable nation.

[vi5]

10. I trust the aggregators......

individual polls not so much.

I don't think we should live or die by them but we can't dismiss them either. That makes us look as loony as the Republicans and their "unskewing".

[PeteSelman]

11. I think they're fairly accurate.

I find it amusing that when Trump is down in the polls, all the Republicans say the polls are fixed or they're phony or they're skewed. But when Trump is up in the polls everybody here says the polls are fake or they're fixed or they're skewed.


Does no one else notice this?

The polls are the polls. If there's an outlier or something that looks ridiculous, fine. But generally they're pretty good.

They can't be just fine when your guy's winning but unbelievable when the other guy is.

[Ace Rothstein]

13. You are correct.

In 2012 the Republicans received some well deserved ribbing for their unskewing of the polls. There was a portion of this place in 2014, probably some of the same people who made fun of Republicans in 2012, who were convinced the polls were skewed/wrong that year.

[Awsi Dooger]

18. Long before that

This site had unskewing obsession in 2004. Months and months of it. Nobody wanted to believe Bush could possibly be reelected. Incumbents whose party has been in power only one term own surreal benefit of a doubt.

[Ace Rothstein]

12. The aggregators are usually pretty close but not perfect.

Nate Silver called 98 or 99 of the 50 states x 2 correctly in 2008 +2012.

[rock]

14. Not much for accuracy

In general, collectively, you can sort of see the trend.

[sofa king]

15. I don't trust them very much, right now.

Right now we're sort of in the same place as we were in 1948, when telephone polls were cheaper to conduct in large numbers but wound up sampling the opinion of a privileged, affluent, conservative segment of America that was not representative of the whole.

Now, land-line telephone polls survey an aging segment of the population that by definition is conservative and not acting in their own best interests (by ditching their expensive landlines and the infuriating telemarketers and pollsters who haunt them). Younger and brighter voters (read: Democrats) don't participate in these polls.

Everyone is politically motivated to misrepresent their numbers in online polls, so they're totally unreliable and probably always will be. Every time I hear, "an NBC/SurveyMonkey poll says," I laugh before I can hear what bullshit it says.

Furthermore, there has been a confluence of interest between polling companies and television news media, both of whom have a strong interest in keeping it a tighter-than-actual race. CNN sold its October ads at top-dollar, so you can bet your ass they were going to make it pay off for their advertisers. Push-polling, which reached its apex in 2012 when the Romney campaign effectively bought out Gallup and manipulated them straight out of their reputation, seems to have gone by the wayside.

The polls that are accurate are the internal polls being conducted primarily by the Democratic Party this year, because those numbers are generated by in-person visits. Trump and the Republicans didn't consider such hard work to be worth paying for.

I think this is why we're seeing the Clinton campaign conduct its business in a way that does not coincide with the fallacious "public opinion" polls which we're all watching. They seem to know what's really going on, at least among their own voters, and they seem pretty damned confident that things are not as desperate as the ad-sellers would have you believe.

[3catwoman3]

16. Seeing as they change almost as often as...

...stock prices, and reported about as often, I am pretty much ignoring them. (Or trying to - kinda hard when almost every other post here is about one poll or another.)

[Awsi Dooger]

17. Averaging plus logical adjustment

This cycle I've always expected a reversion toward tight. Both candidates have low approval and therefore low upside. Hillary was in far better position to win big in 2008, before Benghazi and the email nonsense. A 2016 blowout was not logical given her favorable numbers, although it could have happened if voting had paralleled the release of the bus tape.

Now, I don't trust certain states at all. Alaska has notoriously pathetic polling, and Georgia is weak also. Once Georgia becomes more of a swing state I'm sure the polling will improve as more companies get involved and study the model. It's already more reliable than when I started wagering on politics 20 years ago. Georgia polling was a farce and stayed that way. I don't have similar hope for Alaska. Geography alone makes it a difficult state to survey. Not many companies will make it a priority so number of polls will remain low, and the models have never been decent. Even when Tony Knowles was winning governorship races he didn't threaten the poll numbers.

[jamese777]

19. In 2012

There were many well known national polling organizations whose final polls ended up within a +/-3 point statistical margin of error of the actual Obama 51.1/Romney 46.2 actual election result.
Here's a link to a study of 18 national polls showing that the majority of them called the election accurately in 2012:
http://www.huffingtonpost.com/2012/12/23/2012-polling-accuracy_n_2347772.html

Now just because a poll was accurate in 2012 doesn't mean it will be accurate in 2016. However the worse that national pollsters have ever done collectively since 1968 is miss by 3.1 points in 1980.

[0rganism]

20. watch them for trends, not margins

margins tend to be greatly influenced by the turnout model that the polling outfit applies to the poll, which basically allows the pollster to massage the data in a fairly wide range. for instance, the shift in a recent ABC poll that's freaking everyone out has a "Likely Voter" turnout model based heavily on "enthusiastic support", so if HRC's enthusiastic support decreases slightly while DJT's increases slightly due to external events (e.g. treason by the FBI director), you get an overly large swing in the poll (e.g. HRC +12 --> DJT +1). note that this is independent of the sampling -- complaints about over- and under-sampling of certain groups tend to be spurious although sometimes valid concerns are raised about error margins within specific demographics.

under these conditions you really can't count on the polls for predictive value, but they can show you interesting things about changes in the general outlook for the campaign.

on days like today, however, that information is largely redundant, as the FBI has basically declared war on the Democratic nominee.