Why does the margin of error matter when nearly all polls show JK ahead?

[Fed Up]

I keep seeing these polls that consistantly show Kerry ahead, even if they are within the margin of error. Numbers like 48-45, or 47-44, stuff like that. Yet, if most polling show Kerry ahead, even by 2 points, can't we assume that the margin of error carries less weight in those cases, and Kerry really is ahead?

Say you took 50 polls of A vs. B, and 45 of the polls showed A ahead by 2 points, and 5 showed B ahead by 2 points. All the polls were within the MOE, yet can't it be assumed that A really would win a direct vote? Any statistics majors here?

That aside, I'm starting to like what I'm seeing. Plus a new poll out today shows Kerry ahead of Bush by 10 points in the crucial state of PA, 48-38.

Things are looking good, people, and we're headed into the convention!

[AngryLizard]

1. Yeah, what IS that??

How many of you are tired of hearing, "Kerry is at 47% and Bush is at 45%, so it's a statistical dead heat." Um, no, it means that Kerry is ahead in the polls. Why can't they ever say "Kerry is ahead in the polls"? Now if you want to follow that up by saying, well there is a margin of error... besides which, couldn't the margin of error go the other way? Like Kerry could be at 49 and Bush could at 43, right?

Say it loud and repeat as often as needed: "Kerry is ahead in the polls." Because he is. Consistently. Margin of error be damned.

[qazplm]

2. but, he's not

ahead, they are tied...that's how polling works.

Besides, since polls at this point aren't particularly predictive of anything that happens in November, why stress over it?

The polling to look at is in September and October, THAT is when polling actually starts to reflect what could happen in November because that's when that last large undecided group starts to actually pay attention and pick a team to root for.

[AngryLizard]

4. Maybe so

But I think if Bush was at 47% and Kerry was at 45%, they'd be leading that story by saying that Bush was ahead in the polls. That was mostly my point. The press shirks from ever saying that Bush is behind.


Plus, 2% consistently is still 2%. . . consistently.

[Fed Up]

3. yeh they never say the moe could go the other way. ha. n/t

[AndyTiedye]

5. They Know How the Machinez Work

> they never say the moe could go the other way

Of course not. They know how Diebold Republican Electing Machinez work.

[Name removed]

6. Deleted message

Message removed by moderator. Click here to review the message board rules.

[KurtNYC]

8. Do you want to be right

about all of your nay-saying and cyncism? or do you want to win?

The momentum is against Bush. GOP types have defected in droves. The terror alerts, etc. have lost credibility with people now openly challenging the timing of such alerts.

Kerry's ahead, Bush is freaked out and I'm ready to party!

[Name removed]

9. Deleted message

Message removed by moderator. Click here to review the message board rules.

[AndyTiedye]

18. I want us to win all right.

I want us to win by such a huge landslide that they can't steal it,
big enough so any attempt to stage a "terrorist attack" to remain in power would be
such an obvious fraud that they won't even try it,
big enough so the world can see an unequivocal repudiation of the
Bush Family Evil Empire and PNAC and all that they stand for.

We are ahead a little bit in the polls now. That is good, but
it is not enough of a lead to prevent them from stealing it again.

I don't want us to become complacent, as we all did in 2000.

[Feanorcurufinwe]

10. LOL, those things will happen in Bush's dreams, nowhere else.

Faux focuses on Iran's part in 9/11


What part? :wtf:

According to Faux, we found WMDs about 20 times. Only the brainwashed still believe their lies.

And even it this were true, all it would do is emphasize the fact that Iraq was not a threat and that Bush sent hundreds of Americans to their deaths in vain.

[Name removed]

11. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Feanorcurufinwe]

12. What you don't seem to realize is that all the idiots who watch Faux

are already voting for Bush, so it doesn't matter what they say.


Face up to it, every single trend and leading indicator shows that Bush will lose, and lose badly. Kerry's going to have strong enough coattails for us to win back the Senate and make progress in the House.

[Name removed]

14. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Feanorcurufinwe]

15. I think Abraham Lincoln said it best.

"You can fool some of the people all of the time, and all of the people some of the time, but you can not fool all of the people all of the time."


No matter how much you try to promote the idea that Bush will win, you are wrong.

[Name removed]

16. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Feanorcurufinwe]

17. Goodbye, and good riddance, DaveClearwater99

[KurtNYC]

7. And why do they "weight" the polls now that Kerry

is solidly ahead? CNN admitted to "weighting" their recent polls to skew them towards what they claim are "likely voters." The raw poll of registered voters was like 43 - 47 so they weighted it to make it 45-46.

Seems like step one in election stealing - skew the polls so they will agree with the Diebold results. But everyone knows what is going on so they only hurt their credibility with this stuff. I want to sue CNN for fraud -- this must be stopped.

Kerry is at 47% among likely voters, Bush at 46% and independent candidate Ralph Nader at 4%. Among the larger group of registered voters, Kerry is at 47%, Bush at 43% and Nader at 5%.

So they gave some Nader votes to Bush?! already?!
http://www.usatoday.com/news/politicselections/nation/president/2004-07-22-poll-cover_x.htm

[Still_Loves_John]

13. I agree

I mean, they should still mention that it's within the margin of error, but when pretty much every poll shows Kerry ahead by two points, I assume the lead is real.

[Bill McBlueState]

19. if the methodology is the same...

If every poll had exactly the same methodology, then you could simply add all the results. Instead of 1000 survey respondents, you'd have maybe 10,000 respondents. The larger the sample size, the smaller the statistical margin of error. And the more certain the conclusion that Kerry is ahead.

This assumes the population is large enough that nobody responds to more than one poll.

The problem is, polls ask slightly different questions, select their samples differently, and so on. But it's still a good bet that if every poll is showing Kerry with the lead, he's more likely to actually *have* the lead.

[VirginiaDem]

20. margin of error

Newbie here. I'm an International Studies PhD student. I took an advanced course in statistics last semester. I'll admit it's all still very confusing; it's not my forte. However, I'd have to say that the media misuses the margin of error stat. If Kerry's up by 2 percent and the margin of error is four then of course Bush might really be winning by as much as two percent but he might be losing by as much as six. Everything else equal, chances are Kerry is winning. And, as several posters have pointed out, when current polls show Kerry is up by one or two points then that's the best guess. It's simply incorrect to call it a "statistical dead heat".
Assuming the confidence level is 95% then there is a 95 percent chance that Kerry's "real" margin is between negative two percent and positive four percent. There's a five percent chance that the whole thing is BS. The media, probably to keep it simple or maybe because they're stupid and lazy, ignore the confidence level when they shouldn't be.

Of course, this assumes that the tests are being done correctly and the questions aren't somehow biased.

All in all--Kerry's up by two right here and now.

[Bill McBlueState]

21. confidence level

That raises a good question. For polls, they get the margin of error from the formula

m.o.e. = sqrt(N)/N

For 1000 respondents, that's about 31/1000, or 3%.

That assumes Poisson statistics -- but they never say what the confidence interval is using that formula. You say it's 95%, but it could be as low as 67% or something. I'll see if I have a stats book around....

[Bill McBlueState]

22. 68%

Found the textbook from my undergrad stats class. If the poll result is 48% +/- 2% for Kerry, that means that if you take 100 such polls, about 68 of them will show Kerry with between 46% and 50%.

About 95 of them would show Kerry with between 44% and 52%.

About 99 of the polls would show Kerry with between 42% and 54%.

And there's always the slight possibility that a given poll can be way different from the true number.

[spooky3]

24. ??? I think your book may be talking about something else.

The margin of error is used to estimate the chance that your sample (poll result) accurately captures the population %. If your poll result is 48% +/- 2% MOE for Kerry, and you didn't have other types of errors (e.g., bad questions, non-random sampling), that means that 95% of the time, the actual population % for Kerry will be between your range of 46-50%.

If you repeat a lot of sampling under similar conditions, then your margin of error would go down. If Kerry's sample %s are consistently 2 or 3 points above Bush, the probability is higher (but still is not 100% certain) that he really is 2 or 3 points ahead of Bush in the population.

Please see the layman's language .pdf file I posted a link to elsewhere on this thread. In addition to pp. 10-11, see the earlier pages that lay this out.

[eightyferrettoes]

27. Holy crap, you all

Okay, I just got off of a RW forum (I am turning into a junkie) where I got sucked into a debate with a 50 something guy who could neither spell nor write an actual complete sentence.

I suspect that all this talk of statistics scares Republicans off.

I am so proud to be a Democrat at this moment. *sniff* lol.

[Bill McBlueState]

30. I read the brochure...

What I found and what the brochure says seem to agree on the essential meaning of margin of error. "The chance that your poll accurately captures the population" (what you say) is equivalent to "the chance that other polls will produce the same result" (what I say).

The difference is in the width of the confidence interval that's usually quoted. The brochure says that "48% +/- 2%" gives a 95% (two-sigma) confidence interval. My book says it's a 68% (one-sigma) confidence interval. I hesitate to disagree with either source.... hmmm.

But yeah, as eightyferrettoes said, it's hard to see Freepers having a discussion like this. :D

[TruthIsAll]

32. No it would be 95% prob that Kerry gets between 46-50.

tia

[spooky3]

23. from the American Statistical Association (pdf file)

here is the segment on margin of error--a lot of people misunderstand this:

www.amstat.org/sections/srms/brochures/margin.pdf

See pp. 10-11 in particular.

[CaTeacher]

25. maybe the reason the news media

keep it simple, is because it really IS kind of complicated, and the average sheep would change the channel before they would listen to a huge analysis of statistics?

[newyawker99]

33. Hi VirginiaDem!!

Welcome to DU!! :toast:

[osaMABUSh]

26. We gottem right where want 'em - we're ahead but no one knows

Kerry's ahead but everyone thinks it's tied so that makes the Kerry supporters work harder and Bush thinks he's doing fine.

[farmbo]

28. MOE Elections are much easier to steal...

Eg...with Nader now being on the Reform Party Ballot in Michigan, a relative handful of county...or even Ward...voting officials can flip votes from JK to Nader, keeping *Bush sanitized in the process, and delivering Michigan's electoral votes to Bush.

If we're outside of the MOE in every battleground state, It will take a criminal conspiracy of epic proportions for Bush to get this election before his Supreme Court again.

[Spiffarino]

29. Because it'll look more legit...

...when they steal the election again. That's why they're yammering on about how close it's going to be in November. They want to make the theft seem plausible.

I'm not joking. I think this is their plan.

[Jack Rabbit]

31. Margin of error

In any sample taken from a greater whole, there is always the possibility that the numbers in the sample don't match the figures in the greater whole. If one tosses a coin once, there is a 50% probability that it will turn up heads; however, that doesn't mean that every time one does a series of ten coin tosses that one will always have five heads and five tails.

The margin of error is determined by inverting the square root of the sample size; the greater the size of the sample, the lower will be the margin of error. Obviously, the margin of error on ten coin tosses is very high: about 31.6%.

Tied up with this number is the probability of getting a skewed result. If one makes ten coin tosses, the probability that all ten will turn up heads (or tails) is very low. The probablity is calculated as:

pn

(On edit: that's p raised to the power of n, not p times n.)

where p is the probability of the event occuring on one event and n is the number of events. In the case of the coin toss, the probability of getting heads on any toss is 50%, so the probability of getting all heads on ten tosses is a little less than 0.1%. Of course, 0.1% is also the probability that one will get no heads at all in ten coin tosses.

There are other figures worth knowing, such as a standard deviation. This figure measure how well a given result fits into a bell shaped curve of all results. To find a standard deviation, one needs to know the mean and the variance.

Looking at all the polls, I would conclude that if the election were held today Kerry would win the popular vote by two or three percent over Bush. That would assume that (1) the undecideds would split between Kerry and Bush and (2) that most of these polls are reasonably accurate.