New Time poll

[SackAttack]

I didn't see any threads devoted to this, so I thought I'd toss it out there. Time released another poll yesterday, this one showing Bush with a 52%-41% lead over Kerry.

What's interesting about this poll, compared with the one immediately following the convention, is that the breakdowns seem to be more in line with a properly sampled population - among 'likely' voters, they pollled 34% Republican, 35% Democrat, and 22% independent. Among 'registered' voters, it was 31-32-26.

If memory serves, one of the brouhahas over the post-RNC poll was that Republican respondents grossly outweighed Democratic/Independent respondents, and as such, the sample population was skewed to a point where you'd expect a large lead for Bush.

Given the population breakdown of yesterday's Time poll, what's the likely explanation for such a strong shift in voter sentiment? I'd be interested to hear your thoughts. I'm not particularly interested in dogma from either side of the aisle, as there's a billion or so threads already wandering the forums on the various issues, so please try to check your political animosities at the door.

[JonInMiddleGA]

Given the population breakdown of yesterday's Time poll, what's the likely explanation for such a strong shift in voter sentiment?


-- If the previous poll(s) had a different affilated voter breakdown, then that could account for a shift
-- I'm not sure what definition the Time poll uses for "Independent" voters; self-indentification? The leanings/tendencies of one group of (I) voters vs another that leaned the opposite direction could account for a decent sized shift. As an example, I could be classified as an (I) under some polls, since I am not a "card-carrying" member of either party.
-- Good old fashioned lying on the part of the respondents; maybe about their party affiliation, maybe about their voting intentions.
-- Most likely IMO? Some combination of the three factors above could easily account for at least half of the swing or more. The rest could easily be accounted for by a combination of convention message effectiveness & the absence of any major gaffes at the GOP convention (i.e. the President didn't announce that he regularly had sexual relations with a goat or anything).

While that last bit may draw some scoffs, what I'm getting at is that it's possible that voters previous answering "undecided" were already leaning Bush & were waiting to make sure there was nothing wholly disturbing to them coming out of the convention.

[SackAttack]

-- If the previous poll(s) had a different affilated voter breakdown, then that could account for a shift

It did, and a lot of people felt as though that were the reason for such an inflated lead. Given that the breakdowns are much more in line with the nation's true demographics this time around, I'm curious as to what could account for the shift given that it's got sturdier ground to stand on, this time.

-- I'm not sure what definition the Time poll uses for "Independent" voters; self-indentification? The leanings/tendencies of one group of (I) voters vs another that leaned the opposite direction could account for a decent sized shift. As an example, I could be classified as an (I) under some polls, since I am not a "card-carrying" member of either party.

Sort of the difference between Green Party and Libertarian, then?

[gstelmack]

Given the population breakdown of yesterday's Time poll, what's the likely explanation for such a strong shift in voter sentiment?
I'm one of those people that despite all the statistics classes and population sampling theorems / proofs I've seen, I still believe that if you poll a 1000 people out of a group of 10,000, you still may get the only 1000 in that group that feel a certain way...

And, of course, you're relying on people to give an honest answer. Plus, lots of these are phone polls: how many people actually answer the phone for a telemarketer, or don't hang up as soon as they realise what's going on? Or if it's a street sampling, how many people turned down the chance to be polled? I walk past people with clipboards going "Excuse me!" all the time.

So they get a small sample of people who are actually willing to spend the time to answer their (often poorly-phrased) questions who they are hoping will answer honestly, and use those results to try and sway public opinion.

[CamEdwards]

If I knew the answer, Sack... I'd be making some serious coin working on a campaign.

[Jesse_Ewiak]

Here's a mock-up of the latest polls from pollingreport.com...

http://www.pollingreport.com/images/SEPgen.GIF

[SackAttack]

I'm one of those people that despite all the statistics classes and population sampling theorems / proofs I've seen, I still believe that if you poll a 1000 people out of a group of 10,000, you still may get the only 1000 in that group that feel a certain way...

Yeah, same, but then, I failed statistics twice, so what do I know? This is why I ask for other people's impressions. :)

Jesse, that's even stranger, given the supposedly corrected demographics, that Time should still be showing Bush so much further ahead than the other "likely voter" polls. I wonder what Time is doing that the other polls aren't (or vice versa).

[Jesse_Ewiak]

My personal feeling is that they're just calling JoninMiddleGA 1000 times. :-)

Seriously though, a poll with about 1000 repsondents like most of the polls above will simply be totally out of wack one time out of twenty. Personally, looking at the other polls, Bush is probably up somewhere between two and eight points, closer to eight than two most likely.

[sterlingice]

But, like Al Gore before him, if Bush wins the popular but loses the electoral college, it doesn't amount to a hill of beans (except maybe as some sort of referrendum toward Kerry in office).

From today's news blurb at electoral-vote.com:
"In Rasmussen's 3-day rolling averages, Bush currently leads Kerry by 1.6% nationally, but trails him by 2.5% in the 16 battleground states. Ultimately, the latter number may be more important than the former because rolling up a massive victory in the South and West may be slim consolation for losing the Midwest by 2%. Rasmussen also reports that in the battleground states only 39% think the country is on the right track, vs. 56% who think it is on the wrong track. Numbers like these are never good for an incumbent."

SI

[sterlingice]

My personal feeling is that they're just calling JoninMiddleGA 1000 times. :-)

Seriously though, a poll with about 1000 repsondents like most of the polls above will simply be totally out of wack one time out of twenty. Personally, looking at the other polls, Bush is probably up somewhere between two and eight points, closer to eight than two most likely.
Keep in mind, statistically, what it means when they give you the numbers in a poll:

Poll
Bush 45%
Kerry 42%
4% MoE

What it Really Means
There is a 95% certainty that in a poll of all people, Bush would get between 41% and 49% and Kerry would get between 38% and 46%. So, in reality, it could be 46-41 Kerry or 49-38 Bush. Also, those outcomes are just as likely so the 45/42 doesn't mean much at all since it's within the margin of error.

And that doesn't even take into account that 5% of the time, the poll is completely unreliable and those numbers aren't worth the paper they are printed on. For instance, there was a poll two weeks ago that showed Bush had a 10+ lead in Missouri despite being within a couple of points in every other poll. Another in California by Survey USA had the race within a couple of points depite being 10+ for Kerry in every other poll including those by Survey USA. Sometimes you just do end up with a weird outlier where you end up polling a disproportionate number: if you flip a coin 5 times, there's only a 1 in 32 chance that it ends up heads all 5 times but that *does* happen 1 in 32 times.

SI