Do polls really mean shit?

[Vympel]

Just looking at the polls lately, and I was thinking- how can you possibly figure out which way the voters in general are going to vote based on such miniscule, tiny, infintesimally small samples (oooh, 800 people, big fucking wow)? What I mean is, is it possible that there are a huge pool of voters who have never been queried on their vote? How does polling account for this?

[JME2]

Just looking at the polls lately, and I was thinking- how can you possibly figure out which way the voters in general are going to vote based on such miniscule, tiny, infintesimally small samples (oooh, 800 people, big fucking wow)? What I mean is, is it possible that there are a huge pool of voters who have never been queried on their vote? How does polling account for this?

They don't. The polling is based upon registered voters who have voted in at least one prior election. There's a huge gathering of newly-registered voters out there that aren't eligible yet since they can't vote until Nov. 2. The polls are innacurate and why I don't buy any of this "Bush is leading, Bush is winning" crap.

[SirNitram]

Do you really want multiple years of classes compressed into one post? That tends to lose alot of meaning. But I'll try.

Anything which is effected by many small, random factors will approximately fall into the Normal Distribution(Which is a mound-shaped graph from between 1 and 0 on the Y axis) which describes frequency(The centre of the graph the most common, the ends being the least). The idea is you take a bunch of answers, see how many fall where on the distribution, and extrapolate. There are numerous ways to take into account things you'll miss.. But the trick is to get an accurate sample.

The only way to get an accurate sample of the larger population is for it to be a truly random sample. Gallup's Presidential Exit Poll is an excellent example of this, as they poll every X people emerging from the polls. There's ways to ease this with large, varied group(Called Strata and Cluster sampling, which is basically taking known groupings of people and pulling a 'Simple Random Sample' from each grouping). Here's the thing, though: It must be random, or it's worthless. One oft-cited example is the book 'Women and Love' which sent a survey out to 100,000 women's groups, and got 8% replies. Another one would be website polling. In these cases, only those strongly motivated will get their voice heard, and it's biased towards them.

[SirNitram]

Just looking at the polls lately, and I was thinking- how can you possibly figure out which way the voters in general are going to vote based on such miniscule, tiny, infintesimally small samples (oooh, 800 people, big fucking wow)? What I mean is, is it possible that there are a huge pool of voters who have never been queried on their vote? How does polling account for this?

They don't. The polling is based upon registered voters who have voted in at least one prior election. There's a huge gathering of newly-registered voters out there that aren't eligible yet since they can't vote until Nov. 2. The polls are innacurate and why I don't buy any of this "Bush is leading, Bush is winning" crap.

Those who use cellphones exclusively are also not polled. Throw in that every reliable polling group has the stats so close it's within margin-of-error, and Bush is not winning.

[Fire Fly]

There are large pools of voters that pollsters just simply skip over, such as the youth because tradionally, the youth aren't a major factor in many elections. However, if anything is at all proportional to what's going on at my campus, then there could easily be a significant turn out in the 18-29 category. I can say with confiedence that at least a good 2/3 of the students on my campus will probably be voting. Another factor that excludes the youth vote is that many pollsters do polling via the telephone. I think its safe to say that the average 18-29 year old does not have a regular land line telephone, in fact, its more likely that almost all of them will have a cell phone; I know that each of my friends have cell phones, as do I.

So, how do pollsters account for something like this, for example? I believe I read somewhere that pollsers they don't account for them. Generally when you do a poll, if I remember correctly from my last statistics class, you have to make many, many assumptions. So what happens is that polls tend to only give a general overview of things.

Other factors that are not being factored in is that there have been so many newly registered people and then there will be many more who will do the actual registering on November 2. These people are generally discarded when it comes to polling.

[Kon_El]

The largest use of polls is to tell people runing for president which states they have a chance in and which states to write off. If 1 month from the election polls show that you are 15 points behind in state #1 and 1 point ahead in State #2 you know to cut your losses in #1 and shore up your support in #2 with advertising and personal visits.

Its main use is as a campain tool.

[Durandal]

Just looking at the polls lately, and I was thinking- how can you possibly figure out which way the voters in general are going to vote based on such miniscule, tiny, infintesimally small samples (oooh, 800 people, big fucking wow)? What I mean is, is it possible that there are a huge pool of voters who have never been queried on their vote? How does polling account for this?

Statistical sampling methods is a field in and of itself. Sampling for voters in American elections is a more complex matter.

But based on polling, it's possible to get an estimate for the number of voters in the US. From there, you can calculate the desired N for a simple random sample to achieve a desired margin of error in your prediction. If you poll too many people, you risk corrupting the sample because some portions may not be equally represented in the sample. So you generally stick to the smallest number possible to achieve your desired margin of error.

[The Yosemite Bear]

no

1. they are easy to mess with and slant
2. I refuse to participate in them
3. Even SDN denizens who AIM me have found out that I do not discuss how I am voting in actual elections with anyone....

[Mange]

Just looking at the polls lately, and I was thinking- how can you possibly figure out which way the voters in general are going to vote based on such miniscule, tiny, infintesimally small samples (oooh, 800 people, big fucking wow)? What I mean is, is it possible that there are a huge pool of voters who have never been queried on their vote? How does polling account for this?

When I studied political science at university, we had a sub-course that specifically dealt with quantitative data of this kind. Our lecturers were quite adament that polls are reflecting how people out there in "the real world" thinks at the moment the poll is being made (in other words, a very high reliability). The ideal number of persons to participate in a poll is 1.000 (no matter how big or small the country you live in is). The participants must be of different social background, different geographical locations etc, again in other words, the participants must be representative of the entire population.

Personally, I don't think you should take polls that seriously (especially not those found on the Internet). There are a number of criticisims that can be raised against opinion polls (such as is the selection of participants truly representative?) and I never bought what the 'mathematicians' at university claimed about the infalliabilty of opinion polls (hey, it's only to open a paper the day after an election and see if the latest poll came close or not, many times they don't). I've conducted a number of opinion polls in different topics in connection with my term papers, BSc and BA thesises, but I've always wondered how well the results really have reflected how people in general thinks.

[BoredShirtless]

When I studied political science at university, we had a sub-course that specifically dealt with quantitative data of this kind. Our lecturers were quite adament that polls are reflecting how people out there in "the real world" thinks at the moment the poll is being made (in other words, a very high reliability).

Appeal to Authority; give a logical explaination as to why THESE polls really matter, stop appealing to your bloody education for fucks sakes. You gotta drop that thing down my throat each time you post a friggen N&P post, don't you?

The ideal number of persons to participate in a poll is 1.000 (no matter how big or small the country you live in is).

That's bullshit. The Vatican has a population around 800. And even if it had more, you don't have to have a background in stats to see that the larger the population and the greater its diversity, the larger sample size you'd need. And where did you get 1000 anyway? Is that the MangeSwede Constant I've heard so much about? Or did one of your lecturers say it?

Personally, I don't think you should take polls that seriously (especially not those found on the Internet). There are a number of criticisims that can be raised against opinion polls (such as is the selection of participants truly representative?) and I never bought what the 'mathematicians' at university claimed about the infalliabilty of opinion polls (hey, it's only to open a paper the day after an election and see if the latest poll came close or not, many times they don't).

Quote even one mathematician who said opinion polls are infalliable, and I'll cut off my nuts and mail it to a charity of your choise cause my nuts are so precious and would feed a small country.

[Jalinth]

no

1. they are easy to mess with and slant
2. I refuse to participate in them
3. Even SDN denizens who AIM me have found out that I do not discuss how I am voting in actual elections with anyone....

You've essentially hit big problems that pollsters face. Larger numbers of people are simply refusing to take polls. So this increases the "self-selection" problem - the polled voters might be more partisan than your "typical voter", therefore not representative

Second, how do you define a voter - an expected voter, an eligible voter, aperson on the street? An entire industry exists - the most accurate ones are probably those used by the campaigns themselves. They have the most interest in getting the "right" number and are willing and able to hire the best. They also have very significant private resources (information from volunteers, other sources) coupled with public information to nail down demographic information - as in Group X14 will trend heavily democrat, Group Z91 will trend heavily Republican, Group D19 is variable but doesn't vote, etc... This is not cheap information to gather, but invaluable for close fights.

Third, people like me exist who will (if you hit me in the right mood) will flat out lie. In a small sample, it doesn't take that many to skew the results.

[Kuja]

Yes, polls really do mean shit.

[Mange]

Calm the fucking down, pal.

When I studied political science at university, we had a sub-course that specifically dealt with quantitative data of this kind. Our lecturers were quite adament that polls are reflecting how people out there in "the real world" thinks at the moment the poll is being made (in other words, a very high reliability).

Appeal to Authority; give a logical explaination as to why THESE polls really matter, stop appealing to your bloody education for fucks sakes. You gotta drop that thing down my throat each time you post a friggen N&P post, don't you?

I'm sorry, but I thought that, by the nature of this thread, would be interesting to show how some people in universities etc. believes in the absolute infalliability of what they are teaching. It has nothing to do with me, other than my participation in one such class where I personally witnessed that kind of reasoning. Other than that, I wasn't appealing to my education. Please try to read things in its proper context, I certainly didn't want to offend anyone.

That's bullshit. The Vatican has a population around 800. And even if it had more, you don't have to have a background in stats to see that the larger the population and the greater its diversity, the larger sample size you'd need. And where did you get 1000 anyway? Is that the MangeSwede Constant I've heard so much about? Or did one of your lecturers say it?

The sample of 1000 being "ideal" for any population is what one of my lecturers stated. It was contested during the Q&A, but he was firm. To clarify, it was more or less his own position, but he tried to push it. As for taking the diversity into account, the sample must of course reflect the population in general, but the sample should be reduced to the lowest number possible to avoid statistical errors.

Personally, I don't think you should take polls that seriously (especially not those found on the Internet). There are a number of criticisims that can be raised against opinion polls (such as is the selection of participants truly representative?) and I never bought what the 'mathematicians' at university claimed about the infalliabilty of opinion polls (hey, it's only to open a paper the day after an election and see if the latest poll came close or not, many times they don't).

Quote even one mathematician who said opinion polls are infalliable, and I'll cut off my nuts and mail it to a charity of your choise cause my nuts are so precious and would feed a small country.

That was the funniest thing I read this week (despite the sarcasm). I'm sure they are. You should have noted that I wrote 'mathematicians'. That kind of people like those lecturers I had, are the kind of people that it seems they never steps outside their offices.

[Durandal]

When I studied political science at university, we had a sub-course that specifically dealt with quantitative data of this kind. Our lecturers were quite adament that polls are reflecting how people out there in "the real world" thinks at the moment the poll is being made (in other words, a very high reliability).

Appeal to Authority; give a logical explaination as to why THESE polls really matter, stop appealing to your bloody education for fucks sakes. You gotta drop that thing down my throat each time you post a friggen N&P post, don't you?

Exit polling on election day yields pretty consistent results in terms of networks being able to predict who will get what state. The most notable exception I can think of is Florida in 2000, when everyone said that Gore had it.

[BoredShirtless]

Calm the fucking down, pal.

I can't dude, I have issues.

I'm sorry, but I thought that, by the nature of this thread, would be interesting to show how some people in universities etc. believes in the absolute infalliability of what they are teaching.

You didn't show that but.

It has nothing to do with me, other than my participation in one such class where I personally witnessed that kind of reasoning.

What reasoning? What the hell are you on about?

Other than that, I wasn't appealing to my education. Please try to read things in its proper context, I certainly didn't want to offend anyone.

Too late, you have really pissed me off. Nah just kidding we're friends right?

The sample of 1000 being "ideal" for any population is what one of my lecturers stated. It was contested during the Q&A, but he was firm. To clarify, it was more or less his own position, but he tried to push it.

Just as I thought, your lecturer is a dumbass.

As for taking the diversity into account, the sample must of course reflect the population in general, but the sample should be reduced to the lowest number possible to avoid statistical errors.

Do you know what the "mean" and "standard deviation" means?

That was the funniest thing I read this week (despite the sarcasm).

I wasn?t being funny. I really would have chopped off my nuts if you could have met the challenge.

You should have noted that I wrote 'mathematicians'. That kind of people like those lecturers I had, are the kind of people that it seems they never steps outside their offices.

Bullshit, I don't believe any mathematician even at your Uni would say polls are infalliable. Accurate, ok. But not infalliable.

[Mange]

Ok, BoredShirtless, I hope we're cool and that you understand what I was trying to say in my first post (essentially what you said, that my lecturer was a dumb-ass).

[The Dark]

IIRC, Zogby has called the last four elections (although the 2000 election they said "too close to call." Given that it was decided by <600 votes, and every other poll said big Bush win, I'd say that's a call). This year, they're giving Kerry a very slight edge, but again it's currently too close to call. Zogby himself (as opposed to the Zogby poll) has called it for Kerry based on his Tin Man / Scarecrow poll . Basically, he asks whether a voter would prefer a candidate who's all brain and no heart, or all heart and no brain. Last election that was Gore and Bush respectively, and it was a dead heat. This election, it's Kerry and Bush respectively, and Tin Man was up 9% over Scarecrow.

[Equinox2003]

I personally pay polls no mind. I recall following the 1996 election in
school, with the polls showing Clinton ahead of Dole by 18% the entire
race, we kept track in class. Of course on election night Clinton won by
8%. So much for the 3 point margin of error they always talk about on
TV. anyhow, that is my view on polls.

[HyperionX]

I personally pay polls no mind. I recall following the 1996 election in
school, with the polls showing Clinton ahead of Dole by 18% the entire
race, we kept track in class. Of course on election night Clinton won by
8%. So much for the 3 point margin of error they always talk about on
TV. anyhow, that is my view on polls.

Isn't that the incumbent's rule or something? The incumbent president always gets the same percentage of the vote as people supported in the very last polls. So if Clinton got 49% of the vote, then 49% of the people will vote for him and no more. So if it's like 49-31 it'll end up 49-41 on election day.

[Equinox2003]

Not sure really. My take on 96 is that not enough people had a reason to vote against Clinton, so they did not. Dole did not offer the people enough to win, and therfore, he did not.

[Admiral Valdemar]

This thread would've been far better if Vymp had put a poll up asking the question, oh well.

Statistics is annoying stuff, but thankfully, I had a good lecturer to get me through it since biology practically created modern statistical methods. I certainly don't envy the guys who have to calculate such polls in order to satisfy the public.