Does Anyone Here Think National Polls Have Merit?

[RUMMYisFROSTED]

A recent national poll showed that 60% of Democrats couldn't even name a Dem candidate. That's a poll that you can believe. But why would you believe a national poll, in which presumably over half the respondents don't know WTF they're talking about, that says so-and-so is leading in the polls?

Imo, the national polls won't catch up to public awareness until January or February when the candidates start stumping in more states. Watch the polls that are taking place in areas where the candidates are familiar to the people. Notably, New Hampshire, Iowa and, to a lesser extent, South Carolina.

Am I wrong, or are people just trying to grab on to any bit of good news? :shrug:

[Abe Linkman]

1. Maybe the Pope, but not sure about the rest of 'em

n/t

[RUMMYisFROSTED]

12. The Pope?

(iow, shameless kick.)

[Brian Sweat]

13. Polls - Poles - Poland

It's a bad joke.

[cjbuchanan]

2. None at all

National polls of candidates almost never matter, especially at this point in the game. Now national polls of approval ratings, they are an okay measure of things.

I think you are correct about which polls to watch.

[RUMMYisFROSTED]

4. Thought I was going crazy for a second there.

Until I see the new polls from out of those states, the candidates to watch are still Kerry, Gephardt, Dean and Edwards. Clark hasn't even set foot in those states yet. Those state's voters are extremely serious about who they support. I'm not saying they won't take a liking to Clark but give it a chance to see it play out.


(Listening to a replay of last nights show from Malloy right now. He just said, "It'll all play out in the wash." Amen.)

[DCDemo]

3. Not at all.

I work with statistics and survey data and a lot with national survey of consumer behavior and segmentation of the population into distinct groups of people.

I am NO statistician, but I find fault with the sampling. Non-scientifically, I don't believe that most sample sizes are big enough to realistically represent the US. I don't buy into the Nielson TV data - 30k people supposed to represent almost 300 million?

It gets worse with these polls. The sample is not truly representative -

* people have to be home, this discriminates
* people have to answer the phone, this discriminates
* people have to be willing to take the survey, this discriminates
* people have to have home phone service, this discriminates
* for the most part, people have to know English pretty well..this discriminates

[Larry Gude]

5. DC...

...great points. The same goes for our juries as to the characteristics of who actually sits for days or weeks on end.

More profoundly, the same kind of logic applies to our politicians!

That's why I think we should have terms limits for at least the House of reps. It would actually become the Peoples House.

[DemocratSinceBirth]

6. Polling Is Based On Inferential Statistics

The same method that geneticists use for dna testing....

With random sampling, once you reach a certain number extra samples become superfluous....

Here's a little test to do at home.....

Flip a coin ten times.... You might get eight heads and two tails but flip that coin one thousand times I'll bet you get damn near close to five hundred heads and five hundred tails....

[RUMMYisFROSTED]

7. But what if 60% of the time that you flip it it come up with neither side?

Then how valuable is the data?

[DemocratSinceBirth]

9. Impossible

The more you flip a coin or the more you enlarge your sample the greater your chance of getting an even amount of heads and tails...


That's how those guys get their dna matchs....

When they say it's a 1 in 4,000,000,000,000 matchs do you think they have that many samples....

I think only about 40,000 dna samples exist*...


I have to be careful cuz I am neither a trained statistician or geneticist....

[RUMMYisFROSTED]

19. But a previous poll said that 60% of respondents have no idea

who the choices are. If the sampling ignores those that have no idea and concentrates on those who do, then you'll understand why NH and IA are really the only polls that matter. Follow my thinking?

[DCDemo]

8. Yes, but if you don't know the specs on the coin

You might get more heads or more tails, or even a coin on it's edge (I have seen this once...kinds blows the whole "heads or tails 50%" arguement out the window)

If you measure everyone, you get an accurate answer. If you make a sample that is truly representative and is big enough, then you can make inferences about everyone. If you have a big sample that is not representative, then you get skewed results.

But my point was that these samples are NOT representative. There is bias. Not that I like Dick Morris too much, but he had a good piece on this a couple of months ago. He's a smart guy, kind of a slick opportunist bastard, but smart.

He knows...as do many of the "elder statisticians" I encounter in my work.

Or, to make it another way, do you trust the "call-in" polls, or the ones on major media websites? Of course not - their sample is not representative. There is the same problem with calling people, or, for that matter, mailing to people (I open almost no unexpected/recognized mail for example). Even walking up to people on the street isn't truly representative - some people will ignore you, walk away, etc (not to mention that really affluent people and really poor people would not be as likely to be questioned)

Anyway my point was, these are not 100% believable, and their "plus or minus" adjustment does not truly take into account how biased and non-representative the samples are.

Show me the wording, and the methodology, and I might buy their results a bit more, but anyone can craft a survey that shows whatever they want, no matter what respondees actually say...

[DemocratSinceBirth]

10. We Agree 100%

Polling is an art based on a science...

It's not a laboratory experiment.....


Also, with cell phones, fax machines, caller id, etcetera it's getting harder to get good samples....

That being said... Most of the recognized polls got it right in 00 if you factor in the margin of error....

[Nicholas_J]

11. Well it depends

If you are trying to figure out what the ENTIRE nation thinks, national polls are far more valid than a state poll. After all, we are electing the president of the UNITED STATES, not the president of New Hampshire or Iowa. And even then, the PRIMARIES themselves are not an indication of who will get the nomination. After all Clinton lost both Iowa and New Hampshire in 1992. The last time Jesse Jackson ran he won ALL of the super tuesday states, and we didnt see a President Jackson. All polls do, as James Zogby has pointed out, is indicate trends and directions. Bush is down, dems are up, Clark is up Dean is falling.

Whenever a candidates who is supported by a group falls in the polls, their supporters come out stating that they do not "BELEIVE" in that particular type of poll. But people spend BILLIONS of dollars a year having the same polling companies direct their business decisions and new product development and so on, so I would guess that their techique is FAR more valid than the supporter of the candidates belief.

So yes, within the set parameters of the polls, they are about as good an indication of what the public is thinking as any other tool.

So yes, these polls are far more valid than mere opinion. WHen they poll, they poll those who RECOGNIZE the candidates they are selecting.

[Pepperbelly]

14. It depends upon whether I like the results.

:D

[RUMMYisFROSTED]

15. Now that answer grabs the cat by the tail.

Seriously though, do you actually believe it?

[Pepperbelly]

18. kinda-sorta ...

both ways, whether I like it or not. I wouldn't bet the rent on them but they do raise and lower hopes for us.

[PAMod]

16. They're helpful, you can sort of see how a candidate is being received.

They obviously are helpful when raising money.

Primary momentum is more important than national numbers.

[damnraddem]

17. Of course well-done national polls have merit.

They fairly-accurately portray what a population admits to thinking at a given time. The problem is in interpreting what that might mean. Even a very-accurate national poll taken the day before a prez election is worthless for calling the election unless it is either: (1) overwhelmingly lopsided (in which case there are many other indicators); or (2) can be broken down by state (and no poll has THAT MANY respondents), so that electoral votes can be projected. However, a poll on how people think about the Resident can have a lot of merit, especially when repeated, so that trends can be shown; and especially if verified by other polls that show the same trends, if not exact levels.

[RUMMYisFROSTED]

20. I agree that a national poll has merit when it concerns the

President. Even a dumbshit knows who the President is. But what about a poll in which over half the people polled have no idea what the question even is? That's my point.