Are Polls Worthless At This Point?

[FrankenforMN]

Voter registration has ended in most states, and the already small amount of undecideds have probably made up their minds. In my opinion, all that matters right now is voter turnout. If we get out the vote, WE WIN. It is that simple. I also think that the polls so close to the election are nothing more than media's attempt to discourage voters from turning out, and to get ratings without having to do much work(see CNN/Gallup Poll's rabid inconsistency). Your thoughts?

[IrateCitizen]

1. At this point? That implies they were actually useful at one time. (nt)

[pmbryant]

2. Not worthless. But not very useful, either

I posted a brief "analysis" of the 1996 and 2000 polls, and what lessons I take from their (poor) performance:

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=132&topic_id=1137799&mesg_id=1137799

Peter

[DemocratSinceBirth]

8. In Close Races Polls Lose Their Predictive Value Because Of The M.O.E.

but overall the actual results of most final pre election polls are close to the actual results and by close I mean within the margin of error...

If you like I'll supply documentation ....

[pmbryant]

11. Statistical "margin of error" doesn't explain the bias

If there was only statistical error, polls in 1996 would have had a range from a 4 point Clinton lead to a 12 point Clinton lead, not the 8 to 18 point they showed.

If there was only statistical error, polls in 2000 would have showed anywhere from a 5 point Gore lead to a 3 point Bush lead, instead of the 1 point Gore lead to 13-point Bush lead that they showed.

So while I respect that many individual polls' results are indeed with the margin of error of the final outcome, that explanation cannot explain the 4-5 point bias of the overall distribution of poll results.

--Peter

[DemocratSinceBirth]

12. This Discussion Is Arcane....

We can debate forever about outliers in polling but here's a link of presidential poll performance all the way back to 1936...

http://www.ncpp.org/1936-2000.htm

I'll leave it to others to rate their accuracy...

[pmbryant]

14. Thanks for posting that chart

I've seen that chart before and was looking around for it to include in my other thread on this subject.

I think we agree on the idea that the polls are reasonably accurate. If it's a blowout, they can predict that. If it's close, they can predict that. I'm just arguing that there is a bias that has appeared in at least the last two elections. A bias that is only apparent when looking at several polls conducted by different organizations, and that is comparable to or larger than the statistical margin of error.

So in a close race, the predictive value not only becomes useless, as you state, but becomes misleading.

Using that chart, it appears a similar bias has appeared in many other previous Presidential elections as well:

In 1992, the six polls listed predicted a Clinton win of 6-12 points. He won by 6. Again, at the extreme end of the range.

In 1988, the five polls listed predicted a Bush win of 4-12 points. He won by 7. No clear bias, though 4 of the 5 overestimated the size of Bush's victory.

In 1984, the six polls listed showed a huge range: anywhere from 10 point Reagan win to 27 point. Again, no clear bias, but that range is much larger than the statistical error.

In 1980, the four polls listed all predicted a smaller Reagan victory than he got: from 1 to 6 points, when he won by 10. Not a great performance.

Before that, there are very few polls listed for each election, and it becomes very hard to detect any bias if there was one. No obvious one shows up. (Except for 1948 and 1952 when Gallup was 10 points off each time.)

So the 1992, 1996, and 2000 elections show this bias: where the final outcome was only correctly predicted by an "outlying" poll.

Before that, the track record is somewhat better (except for 1980), but that was a very different polling universe than exists today.

So I have every reason to believe that this bias still exists today.

But I also have every reason to believe that, indeed, this is a close race.

--Peter

[papau]

3. No - indeed they indicate a Kerry victory!

[rocktivity]

4. You're absolutely right

Especially if early voting turnouts are any indication. For instance, check out this DU thread about what's happening in http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=180&topic_id=5608">Texas.

:headbang:
rocknation

[Demfromct]

5. I agree

The race is Bush up between 1-3 points the way I see it before undecides. It think the have made a decision but haven't told people. The tracking polls with the excpetion of Wahsington Post are in this range. I think the post will come back to three points.

Bush is only above 50 percent in one poll. Kerry is between 45-47 points. If he can get those undecideds to break for him in a high turnouyt election he wins. If he can't GOTV in Wisconsin and OHio, it is over. I think Bush is in better position than last week, but it is still fluid in my mind. Slate's analysis of the Mason-Dixon polls in their election scorecard helped ease me. The Michigan and Wisconsin surveys were whacked out. And the Mason-Dixon poll only only polls about 650 people. I know how the stats work, but I still think that is to small. There is so many different dynamics in this race I don't know how u can cal 650 people and get a good enough sample.

[DemocratSinceBirth]

7. You're Agreeing Polls Are Worthless But Your Whole Analysis Relys On Polls

nt

[ClassWarrior]

6. YES!!!

That is all. Carry on.

24.

[Nicholas_J]

9. Close to useless

Two out of the latest three national polls show Kerry head of BUsh the other shows them even. Then go to national tracking polls, and the numbers are reversed. Though the state polls in this case are more valid than the national polls, we are gettinf the same things. Many polls show Kerry ahead in the battleground states, but a few show the opposite, whith Bush having major leads in states where every other poll for those states showing Kerry ahead. Example for the last week of so, almost every poll shows Kerry leading in Ohio. Faux of course must then do a poll showing Bush 6 points ahead.

I thinks its about time for pollsters to be licensed, and national standards set up each years for them to do their polls by.

This time around it seems like all but a few polls are being used as intimidation tools, in an attempt to discourage voters, rather than being used as tools to figure out trends. If you excluded all polls that show some sort of bias towards one party or the other, almost all polls would have Kerry with a tiny lead or in a dead heat with the president. Gallop of course is one pollster who is oversampling Republican, so much so that bookies have discounted it in theiri calculations as to who will win, and among betters, Kerry has the lead for the money. But other pollsters seem to be doing the same. SUSA is oversampling Republicans as well, and so is Harris.

Very few pollsters are actually sampling voters either equally, or in proportion to their historical turnouts. Zogby does, Ipsos Cook does (they are probably the worlds largest pollster, with earnings of something like three billion dollars a year, they are in competition with major oil industries for income) ARG does. I wish some of the pollsters who polled during the run for the nomination were polling, Like Suffolk U, as they were not only accurate, they published figures of virtually every type of breakdown of voters sampled.

Yes, it seems that conservatives can corrupt everything, and after this election, polls are going to be regarded with a good deal more skepticism than they had been in the past. Zogby and one or two others seemed concerned enough for the reputation of their field to very vocally point out the misuse of polling data and models this time around, going so far as to poll to see if people beleived the presidential polls. it seems they dont.

[DemocratSinceBirth]

10. There Are Standards..

check this out.


www.ncpp.org

[Nicholas_J]

15. Unfortunately

They are either too vague to be useful, or they are not requireed to be followed. For example Gallops statements that they beleive that 24 percent more Republicans are going to vote in this
election than Democrats when all data from past election, and past recent presidential elections do not indicate anything of the sort, How one can use the same standards to get such wildly differing results is an indications that either the rules are incorrect, or that these standards are being ignored. How ARG can show that 38 percent democrats will vote compared to 35 perccent Republicans, but Gallop can get 39 percent REpublicans being likely to vote and 30 percent Democrats is a rather large variance. Much too large to reflect actual conditions.

[kostya]

13. What are ALL of you doing about GOTV. That's what matters now, not

some stinkin' poll. It's basically TIED, so forget about it and work a swing state either there or from your state on the phone on the internet, write LTTEs, whatever! Big Dem turnout is crucial, but it doesn't just happen y'know? Kerry On! - K

[pmbryant]

16. Don't worry. We're working hard.

I'm on a project to get Dems out to vote early here in Texas. (Ok, it's not a swing state, but it's where I live.)

--Peter

[Roland99]

17. YES! Stop posting about every frickin' poll that gets released!!