Are the Zogby Interactive polls reliable? Which polls are reliable?

[fujiyama]

Are they really very reliable? They have a self selected pool of responders, so is it reliable if it isn't random?

I'm just wondering because while I like the Zogby numbers (they are usually somewhat favorable for Kerry), I'm not sure if they are accurate.

What polling firms usually are accurate?

This is the impression I've got (correct me if I'm wrong):

Fox News polls seem to be very unreliable and very favorable for conservatives (no surprise considering FOX News itself is propaganda). I think the LA Times polls are usually favorable for democrats and liberals (though I don't think they are as off as Fox News), so I'm not sure if they are really accurate either. I also remember they had some polls showing Grey Davis with a better chance than he did have. I heard Rasmussen is a right winger and he was pretty far off in '00, but his numbers are somewhat good for Kerry this time -- at least his state by state polls are. ARG and Survey USA have been pretty good overall (though Survey USA was really far off in a CA poll a while back showing it be a dead heat). Quinipiac seems somewhat good for conservatives. I'm not sure about Mason Dixon though.

As for national polls, I think Gallup is usually slanted to the right. That NC poll really didn't make too much sense. I'm not sure about Time, CBS, NYTimes, NBC, Washington post and the rest....There could be a bias, but I haven't identified them yet.

[Eric J in MN]

1. No. Use Zogby's normal polls, not the interactive ones.

Interactive (internet polls) are just for entertainment.

Also, check out the CBS/NY Times poll on the right of this page:
http://www.nytimes.com/politics/

[democraticinsurgent]

23. zogby 's interactive is the future of polling

i don't agree that it's just for entertainment. Zogby, like most pollsters, know that phone polling is endangered by cell phones. how many people do you know that have dumped their land lines and gone all-cell? anyone who does this becomes invisible to pollsters.

one of the pollsters that NPR talks to regularly said last week that one of these days, the whole (phone) methodology is going to collapse. My guess is that's sooner not later. Like by the 2006 elections?

look at zogby's web site and he does a good job of explaining how they are exploring interactive as a new method and they know it has bugs still. i applaud him for looking to the future.

Yet--to your specific point--the current Zog interatives do have to be viewed with caution.

[louis c]

2. I Rely on Rasmussen

I read them all, however, rasmussen polls 500 new people every day.

Tracking polls factor out a daily volatility. For instance, you could see the ebb and flow on Bush's bumps around Memorial Day and Reagan's death. Then the polls settled back. Same is happening to Kerry right now.

Rasmussen gives up to date info, and he tends to be a right winger.
This being the case, any Kerry leads would really take on the air of credibility.

Zogby is fine, and I always like Gallup. But those two are just a snapshot, and not too recent at that. Putting them all together can give you a trend, but I check in with Rasmussen every day.

[aquart]

3. I take Zogby email polls.

I signed up in response to a request from Buzzflash, which worked with Zogby to give them a greater Democratic representation.

[tritsofme]

4. Zogby Interactive polls are worthless

They aren't scientific, and they aren't worth your time, regardless of who is winning.

I'm very suspicious of Zogby state polls as well. Although he nailed the popular vote in 2000, however he blew the polls in the 2002 midterms and the Dem primaries.

Survey USA wasn't around in 2000 IIRC, so all we to judge them by is the California Recall, which they nailed right on the head. That poll that have Bush even with Kerry in CA was most likely an outlier, and I wouldn't discredit them based on that alone. I think these guys will be very reliable.

ARG blew alot of states in 2000, and I watch them with a cautious eye.

My favorite state polling firm is Mason/Dixon. They have a good track record and I trust them greatly.

Rasmussen, while it is good for looking at trends has questionable reliability. He blew 2000 in a bad way toward bush, and I don't trust them.

These are my favorite polling firms:

1 Mason Dixon
2 Survey USA
3 Research 2000
4 ARG
5 Rasmussen
6 Zogby

On the national stage my favorite is Gallup. They have a proven track record of over 60 years, and I see them as the gold standard of national polling.

I generally disregard Fox/Opinion Dynamics, LA Times, and to a lesser extent CBS.

It has been my contention however, that especially in national and swing states, that as the election nears polls will be basically worthless because all of the results will be within the MoE.

[TruthIsAll]

11. Wrong. Zogby had it right in 2002. The Repukes stole the senate.

So of course his polls were off. He didn't factor in the fraud.

Here is my probability analysis model which provides circumstantial proof that four states were stolen in which the Dems has leads beyond the MoE in the final polls.


http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=104&topic_id=85732

[tritsofme]

12. Sure boss

If it were that simple every single Democrat elected official in the country would be crying it from the rooftops.

[ISUGRADIA]

14. You still miss one of the basic tenants of statistics

Polls are just a sampling of an exact period, a snapshot of the moment in time. So you cannot say that final results prove or infer fraud based on dated polls. M of E is for that poll for that time period only.

Unless you use sampling the day of the election, you don't have the exact universe to poll from, only self-identified voters or likely voters.

Plus, what about the survey samples? If a survey includes 30% African Americans and only 22% show up, of course results on election day will be different. If a pollster has used an inaccurate sampling base, of course results may be different.

And there were other polls that showed much different results for the four races you cited, some from the same firm.

Several polls in Texas had Cornyn ahead by sizable margains, Sunnunu was ahead in final polls in NH, The GA polls had it neck and neck to the end, and MN was a bizarre election that had a lot of last minute variables that effected voter choises.

2002 Poll Source:

http://www.realclearpolitics.com/Congressional/Senate_02_Polls.html

[depakid]

15. Randomization and a large enough clustered sample

should cancel most of the confounding variables out- and there are statistical methods that can be applied to show that polls taken very close to an election should be accurate within their margin of error to a certain confidence level, usually 95%. I say should, because polls often end up with bias (like non-response bias) in their sampling- intentionally or unintentionally. When there's bias in the sample, all bets are off.

You are right on the mark about exit polling- although since we didn't have any nationwide exit polling in 2002, we can't use them to demonstrate DRE fraud in the Georgia election. Even so, the poll swings from several days before the election were dramatic and more than suspicious.

For example:

Atlanta Journal Constitution Poll Nov. 1

Roy Barnes (D) 51%
Sonny Perdue (R) 40%

Diebold results Nov. 5

Roy Barnes (D) 46%
Sonny Perdue (R) 51%

Atlanta Journal Constitution Poll Nov. 1

Max Cleland (D) 49%
Saxby Chambliss (R) 44%

Diebold results Nov. 5

Max Cleland (D) 46%
Saxby Chambliss 53%

Sample size was 808 but I couldn't find the exact margin of error for those numbers- I think it was 4 points, which would make swings of this magnitude over one weekend highly improbable.

[DemocratSinceBirth]

17. The Confidence Level

Even at a 95% Confidence Level there is a 5% chance your poll is totally off...

[depakid]

18. Actually

all it means is that there is a 5% chance (P value of .05) that the poll will fall outside the plus or minus 4 range. It doesn't mean the poll will be totally off, 4.5 would be enough, and the farther outside the range the actual numbers fall, the less the likelyhood is that it's due to chance error.

Of course, to be sure, you'd have to know a lot more about this particular poll, but my understanding is that it's fairly close to what was reported in other polls taken around the same time.

[DemocratSinceBirth]

5. Rasmussen Is A Joke

That's why you don't see his polls included at Polling Report.Com.
He uses an automated caller. You give your choice to a damn machine.

As for the flawed Gallup NC polls all polls have a confidence level of 95% as well as a margin of error.

With a confidence level of 95% that means there is a 5% chance that your poll is not measuring what it intends to measure...

Do a poll of polls or throw out the outliers.... The highest and lowest for your candidates....

Despite the popular impression Zogby wasn't the only pollster to get it right in 00... Harris, CBS, and Newsweek nailed it in their election eve poll. They called it a draw or a ever so small Gore lead...

[democraticinsurgent]

9. automated polls not necessarily bad

there's a pollster i know here in Indiana who uses the automated methodology. the old-schoolers have been hammering him for years, but election after election, his polls have been more accurate.

The live phone-outs have bias introduced by the callers, while the automated method has the same voice on every call.

On the surface, the automated methodology should not work so well. But in this local case, it's been uncannily accurate, even forecasting upsets and leading trends.

[Killarney]

6. In 2000

I think Zogby was pretty close. Gallup was way off.
Rasmussen has a right-bias.

[tritsofme]

7. In 2000

Zogby nailed it.

Gallup got Bush 48 Gore 46 Nader 4.

And Rasmussen has something like 7 point advantage for bush.

[spooky3]

8. don't forget to check the trends at Pollkatz

http://www.pollkatz.homestead.com/



"Which pollsters like Bush the best?"

http://www.pollkatz.homestead.com/files/bushindexprobushtable.htm

[DemocratSinceBirth]

10. Here's Why Rasmussen Is Bullshit...

The generic ballot on 7-11-04 was 46% D 38% R

Today it's tied at 41%.....

No way is party identification that fluid...


If you don't have any idea of how many Republicans, Democrats, and Independents are in your sample how can you have a credible poll...

[YellowDawgDemocrat]

13. I just look for trends

I don't put any faith in any of the polling people. I simply look for recent trends and consider the current news to get a pulse on the public. That's about as good as you are going to get.

[DemocratSinceBirth]

16. Just Do A Poll Of Polls Or Throw Out The Outliers...

This will give you a good idea of what's going on...

[depakid]

19. but what if the outlier contains the most accurate results?

You'd have to know more about the methdology of the individual polls to do that, as most of these (Rasmussen excepted) tend to be fairly close together and so aren't really outliers in the technical sense.

I tend to agree with yellowdawg, political polls may be useful in identifying trends, but have little predictive value unless they're conducted very close to or on election day. More often they're used as persuasive rather than descriptive tools.

[DemocratSinceBirth]

20. Polls In Campaigns Are Mere Snapshots

but they they let you know how you are doing...

[DemocratSinceBirth]

21. The Outliers Can't Both Be Right...

For instance if one poll had Kerry up by ten and another poll had Kerry down by ten and they were taken on the same day from the same population one is wrong...

[Feanorcurufinwe]

22. "self selected pool of responders" ?? Where did you get that from? nt