Karl`s PC Help Forums

New Hampshire--polls failed to predict Hillary Clinton's win
scholar - 9-1-2008 at 11:23

article and reactions to article

All the polls made wrong predictions, thinking Obama would win.

The article seems measured and reasonable to me.

The reactions/comments show the extreme reactions in some people's thinking, including:
1--Vote fraud by the Clintons, to get a faked win
2--Vote fraud by the Republicans, giving a fake win to Hillary because they would rather run against her.

Those people have more confidence in the polling data (often done by telephone) than they do in the actual votes!

Yikes!shocked_yellow


scholar - 9-1-2008 at 16:12

After I made my previous post here, it occurred to me that some comments about the nature of polling are in order, for those not familiar with the mathematical science behind it.

Technically, the polls always qualify themselves in such a way that they are technically not wrong, even if their picture does not match what happens. When polling is done, a sample is taken, but there is always the possibility that the sample (even if it is random) does not exactly reflect the larger group which one wishes to predict. An example you might imagine in your mind, suppose there are one million marbles, of which only 5 are green. If you reach in your hand blindly and grab 5 marbles, it isn't liikely that all of them will be green--but it is possible.

A pollster is looking at the problem from another approach: if he draws 5 marbles out of a million and they are all green, what is the most likely color make-up of the million? Your own first intuition might say, "If every marble I've drawn has been green, five in a row, I'm getting the feeling they may ALL be green."

Mathematicians have formulae which not only make predictions, but which also match the predictions with levels of confidence, based on the size of the sample, the size of the larger group from which the sample was taken, etc. If you want to be 90% sure that your prediction is right, you have to include a certain number of answers. If you want to be 95% sure your prediction is right, you have to include a larger number of answers (because you had to add some possible answers to be more sure you hit the answer right).

The actual data of the polling companies includes all the details--the size of the sample, the level of confidene (whether it is 95%, or higher, or lower), and the margin of error (usuall expressed as + or - a certain number of points, which is the number of different answers for the specified degree of confidence.

So, the improbable answer is covered. If the pollster says the sample tells him you can be 95% certain that candidate A will get x votes, plus or minus y points, he is also saying there is a 5% chance that the answer will be outside his prediction.

All this is quite apart from other complications, such as the fact that the survey sample is for the time at which it is taken. If people change their opinions because of something that happens the day after the opinion survey is taken, it does not take account of that effect.:)

It isn't necessary to assume that someone has tampered with the votes.


crikey - 9-1-2008 at 19:56

Chads had their day then, Scholar? ;)


scholar - 9-1-2008 at 20:24

I don't know if anyone is still using that system.

The people who believe there is vote-tampering now think that the electronic voting machines are either in the control of people who preset them with lots of votes for their candidate, or they think they are hackable.

There are some advocates of electronic machines also making a physical record, which could be checked against the electronic total in case there was some question about hacking. However, that takes us back to the paper ballot problem--how can tens of millions of paper ballots be counted quickly enough?


Redwolf5150 - 10-1-2008 at 17:32

There is a third option:

Voters deliberately lied to the pollsters outside the polling places.

While I work for a newspaper, I have never liked exit polls. The few times I have been polled by someone outside the voting place, I have told them I voted for the candidate I didn't vote for.

It doesn't matter what the pollsters say anyway. The only thing that matters is the actual ballots cast by the voters.

kewl_glasses


scholar - 10-1-2008 at 18:11

When I first posted in this topic, I was mainly concerned with the predictive polls, but then took up the exit polls as well.

Redwolf, do you think it likely that people were lying during the telephone polling in advance of the voting?

I heard some pundits on discussion shows suggest that people changed their minds after the predictive polling stopped, but before the vote (that is, during the last two days). However, in the exit polls, Hillary led Obama among voters who said they had made up their mind in the last few days AND among those who said they hadmade up their mind earlier.

The ABC polling expert said that he thinks getting the sample from those who would actually vote was the problem. When you are taking the predictive poll, you don't know for sure if you are talking to people who will actually vote, or not. They may say they will vote, and then not. Or they may show up to vote for the first time time, where the models suggest that most people who have skipped voting before are likely to skip it this time as well.

Part of Obama's big win in Iowa was a large turnout of young voters. Three college towns in New Hampshire did not have similar large turnouts.

I have wondered if the expectation that Obama would do well worked against him. If your guy is the underdog, you're highly motivated to vote. If you think he will win easily, and your vote will just add .002 to his margin of victory, you might not want to drive over to vote.


Redwolf5150 - 10-1-2008 at 18:35

Having taken courses in college to learn how to conduct polls and surveys, I can tell you that they don't have the "margin of error" included for no reason.

In any poll, people are going to give bogus answers. some people are just like that.

kewl_glasses


crikey - 10-1-2008 at 19:09

Quote:
Originally posted by scholar


Part of Obama's big win in Iowa was a large turnout of young voters. Three college towns in New Hampshire did not have similar large turnouts.



And, as our papers are saying, that might well have been due to the fact that large numbers of college students who attended Obama's rallies may not have been registered to vote in New Hampshire? I'd have thought pollsters would be aware of that?

Another suggestion, which I find equally suspect, is that the good folk of NH would be reluctant to tell pollsters that they did not support a black person, so they lied.

Whatever, I'm inclined to switch off completely until the contest is settled with the final counts!


SRD - 12-1-2008 at 23:01

Being a citizen of the UK I hesitate to dip my toe in here, but from where I'm sitting I don't think Mrs Clinton should be resting on her laurels, the result was quite close, especially for a state where she was expected to win easily. I think the polls pointed to this, there was a swell in favour of Mr Obama that may have suggested a vast swing, but surely that big a swing was unlikely and so the poll should have been considered in this light as a rogue poll rather than an indictment of the whole system or as a true pointer as to how the progress of either candidate will continue. I think the result shows that there is a very close competition coming up and when results are close polls often get things wrong.


scholar - 12-1-2008 at 23:17

SRD, the predictive polls which I had in mind, in posting this thread, were only about New Hampshire (not drawing conclusions about what it would mean for any other state). They were 9 in number, and all agreed that Obama was expected to win by a comfortable margin (most predicted he would win by eight or nine points, some predicted it would be as little as 5, I think).

That's what drew such interest--not even one poll agreed with the actual result, and it was outside the margin of error for the customary confidence level.

I think either Redwolf is correct in thinking people did not answer honestly, or there was an error in predicting who would actually vote.

You may very well be right, that it could be a long and close contest. I think the longer the outcome is in doubt, the more harsh the battle will be.


liz - 12-1-2008 at 23:36

To be honest, I don't pay attention until it's time to REALLY vote for the final candidates. :) I don't pay attention to exit polls or news reports about surveying the general public.

To me, what counts is the actual number of votes. I just wish that is how it works in the end.


marymary100 - 12-1-2008 at 23:40

It's going to be a long drawn out process.............again.


scholar - 13-1-2008 at 20:42

It's good to see you posting again, Liz! :Dwaveysmiley