Those polls they like to throw in the headlines every week showing whose ahead are pretty much bullshit aren’t they?
Todays shows McCain with a now 7-point lead over Obama from likely voters.
I can’t believe people just flip-flop on a weekly basis based on ads runs on tv.
I’d think it’s more likely the random sample of those polled changes every time and is just too small. Call 5000 random people and McCain has a 7-point lead. Call another 5000 and Obama has a 7-point lead.
I’m just not buying these things as being as accurate as they’d like you to believe.
I’ve seen today some stories that read McCain has the lead, while other stories show that Obama’s lead has merely been cut to 3 points.
CNN can’t even get thier polls straight. Their poll of poll graphic on their Politics page shows Obama with a 1-point lead (45-44), but this story says the CNN poll of polls has Obama up by three points.
Then there’s the new Zogby/Reuters poll that has McCain up by 5 points.
I’d say, if anything, because of McCain’s strong attacks of Obama recently, and Obama’s recent vacation, we should just call it even right now and go have lunch. These nickel and dime ups and downs don’t mean shit, especially in August.
Huh, Obama’s still in the lead according to two of those three cites I provided.
He could have a 10 point lead right now, and I’d still think they’re bullshit.
It’s the nickel and dime up and down that doesn’t mean a hill of beans no matter who’s up-- “Ooh, he’s up 3 points today!” “Wow, he’s closed the gap to 2 points!” “Look at how he’s down 5 points this week!”
Funny how you make unsubstantiated generalizations with nothing to actually contribute to the discussion…
I agree that it is not useful to be obsessing about poll numbers this far out, they do tend to fluctuate a lot. For the geeks, realclearpolitics.com has a good, frequently updated list of polls. They are arranged by State, allowing you to see how the electoral vote is trending, which of course is all that really matters (they have Obama comfortably but not decisively ahead). But to the OP, I have some training in statistics, and assuming that the questions are decently phrased, and that the people being polled are actually a random sampling of the population (neither of which are necessarily safe assumptions, or easy to evaluate based on news reports), 5000 people, or even 1000, should be more than enough to get a highly reliable result; so if the polls are fluctuating that wildly, either people are really highly indecisive or the polls are of varying quality. I suspect it is more the latter.
A poll is only meaningful if it is neutrally administered and repeated.
If I want to show a poll is FOR a candidate it is easy to construct a series of questions that favor that candidate. I could even pick the questions as a subset and present them as a separate poll. By that I mean I can ask someone to rate each candidate on a series of 50 questions, and then pull 10 of them out favoring a candidate. How many times have you seen a “news report” that says something like this: “when asked how the United States should handle illegal Chinese cricket racing 70 % said that candidate “A” would be the best candidate to deal with the crisis”. The implication is that 70% support that candidate. It doesn’t matter if illegal Chinese cricket racing is low on the list of concerns.
If, however, I want to accurately predict public opinion then I would poll people to find out their greatest concerns, and then construct a generic poll repeated over time. By using the same questions it has the effect of expanding the original sample size and also takes current public opinion into consideration.
Are you looking for a semi-accurate predictor of the future or are you looking for a semi-accurate recap of people’s desire? They aren’t the same thing.
As a predictor, I would probably look at market based systems or betting because I think people are more careful with their money than they are with their opinions.
Could McCain be seeing a bounce due to the forum he and Obama participated in over the weekend w/ Rick Warren?
But National polls don’t tell the story like state-by-state Electoral College polls do. At any rate, I really don’t think these polls mean much until at least after Labor Day. It’s better to be up than down, that’s for sure, but things are too changeable.
If the sample is even random. The university I went to often conducts political polls. And they do it largely by polling students at the university. How meaningful is a poll if the majority of those polled are 18-22? It was neat to be involved when they’d call our dorm rooms for the polls, but I doubt the average college student and the average mid-aged or elderly voter in the state are concerned with the same issues.
And speaking of meaningful - do any of the polls restrict themselves to polling people who have actually voted in the past (or have registered to vote if too young to have voted in the past)? Not just “likely” to vote, but actually have gotten to the polls at least once to cast a ballot. I seem to recall polls where Gore and Kerry had commanding leads over Bush, but when all was said and done a lot of their supposed supporters didn’t bother to vote.
Polls by research firms who are paid to conduct such polls are very statistically reliable. However, in saying that, I mean (and they make the disclaimer) that the polls are only accurate to a specific confidence level. If you read the fine print you’ll see that the poll is accurate 19 times out of 20, or to a 95% confidence level, for example. The math is proven. Now, if you ask completely different questions, then yes, you may get completely different results. Independent research polls are a heck of a lot more reliable than you think; they’re based on proven statistics.
Polls showing who’s ahead by national popular vote are B.S. by definition, yes. We don’t elect Presidents that way. Gotta add up the state polls by their electoral votes to get anything useful.
That said, if the margin is just a few points either way in every poll, even with their varying data-reduction methodologies, it does tell you the race is pretty close. Or it would be, if it worked that way.
As Leaffan said, the poles from the professional polling firms are generally statistically sound. They attempt to take proper random sampling and ask non-biased questions and all the rest.
However, where they can be misleading or wrong is when they apply correction factors based on the observed demographic makeup of the people they polled. For example, they have to correct for what they perceive to be the political distribution for the population as a whole. If the country as a whole is 60% democrat, 40% republican, then they’d want to pull their polling numbers randomly from a sampling of those proportions. If the country is actually 50/50, then the answers will be skewed towards the Democrats. Or skewed in favor of the Republicans if Democrats have even moe popularity.
But the big deal with reading the polls right now is that most Americans haven’t started paying attention, so their answers are based on little information or interest. After labor day, and people start listening to them and watching them debate, the numbers will change.
I think you’re likely to see a repeat of Reagan’s election here, with Obama in the Reagan role. In that election, It was very close right up to the last debate, then Reagan exploded and won by a landslide. What appears to have happened is that Reagan’s ideas were popular, but the country was nervous about Reagan the candidate. They worried about his age, they worried about whether he could handle the job. The last debate, when Reagan blew away Carter, seemed to flip a switch. At that moment, Reagan looked like a president. The fears of his being up to the job vanished, and the rest is history.
People are nervous about Obama. They like what they see, and they are impressed by the oratory, and they might even like his ideas. But he’s a risky pick. He’s a green politician who hit a golden moment in time. His background is unconventional, his candidacy is unconventional. I’m sure that’s why he picked Biden - he realized he needed to ground the whole thing a bit and make it look more pragmatic.
The debates are going to be key to this election. Until then, it’s a crapshoot.
Sure the math is proven, but you’re missing something. The poll make be accurate 19 times out of 20, but it’s only accurate within a certain range. The common phrasing is like “poll accurate within 2.5 percentage points 19 times out of 20”. If you poll fewer people, the range is wider (i.e. within 5 percentage points 19 times out of 20). One could conduct a poll of Friday morning showing Obama with a 2 point lead and one on Friday afternoon showing McCain with a 2-point lead (OMG! A four-point swing!) and they’d both be accurate, or at least as accurate as polling ever gets.
