If you were a bookie... Polls and bets on the 2004 elections

Now, that's lovely news indeed.

I guess we have a disagreement about what constitutes a Convention bounce...

My understanding has always been--a bounce is sort of an emotional and IMMEDIATE swell of support for a candidate, post-Convention. It is usually a spike of 8 or so percentage points, which can dissipate or climb--based on othrer, later factors.

But, you can't call percentage changes part of the bounce after about 48 hours...

This is why CNN, USA TODAY and EVEN CBS headlined with No Bounce For Kerry. They, and many others (me, too) consider the bounce to show up soon after the Convention.

The CBS poll nimh's article above alludes to headlined: No Bounce For Kerry--but went on to say Kerry had made inroads with vets.

They say he got no bounce. I say he got no bounce.

Bounce isn't the be all, end all--but shouldn't we at least agree whethrer or not he got one? What are others' critieria? Mine is poll changes within 48 hours of the last Convention day.

Well, Sofia, after all the conventions I've seen, the bounce was talked about the day after the convention.

The bounce comes the day after the convention, and generally goes down over a course of 10 days to two weeks. That doesn't mean the candidate cannot have a lead after 10 days, only that usually his lead is back to where it was just before the convention.

If Kerry increases his lead over the next two weeks, that would be unusual. And very bad news for Bush, since it would seem to indicate the electorate just got to know Kerry at the convention, and likes what it sees the more it sees him.

However, the usual thing is for the bounce to last 10 days-largest the day after the convention, slowly dissipating over 10 days to two weeks.

The Brand X poll:

49% hate Bush
50% like Bush
1% like Kerry

Ehmm ... look again at the list I posted.

All the polls listed there WERE done within some 48 hours after Kerry's speech.

Kerry's speech was on the 29th, in the evening -- and all these polls were done between July 29 and August 1.

So I dont really see what your point is, here.

9 out of 10 MD's don't drive Chevrolets

9 out of 10 Proctologist drive Probe's

To answer in more detail: most pollsters simply dont poll every week, let alone every day. Most only poll once every two or three weeks. Often, they poll in the weekend. "The weekend after" has thus been the usual time to check the post-Convention (or any other) bounce. And a "bounce" is measured by the size of the swing: the number of points the margin one is leading the other by has changed.

Now trying to measure to what extent there was a "bounce" after this Convention, I merely listed the change that occurred in each respective poll in the percentage difference between the two candidates, from one poll to the next (I'll do ya a favour and change the word "bounce" to "horserace change").

Now if that change is negligeable (or even negative, as in the Gallup poll), then there obviously has not been a "bounce", or at most, well: a negligeable "bounce" ("baby bounce", Newsweek called the 2-4% bounce its poll had measured).

In this case, the average "bounce" registered in all the immediate post-Convention polls was all of 2% ... "negligeable", as I already noted. Less than a baby bounce.

----

Now to me, looking at all these polls and the bounce (or lack of one) they purport to show, the problem is on the other end - but not a single media has picked up on it, so I guess its just me.

It's this: all of these polls DID poll right after the Convention - but most of 'em hadn't polled right before the Convention. Because of the above-mentioned reason: they simply dont poll every week.

So they're measuring the bounce based on numbers from, indeed, within 1-3 days after Kerry's speech - but compared to numbers from sometimes quite a while before the Convention.

In my view - but again, I havent heard anyone else comment on it - that runs the risk of not being able to isolate the effect of the Convention from any other things that may have been taking place in the week or two or three before the Convention started.

Hence my theory that it might be possible that there was in fact a (small) Convention bounce, but that all it did was merely cancel out a Bush bounce that took place the week or two before. That's mere speculation, though.

If you're looking for polls that actually compare the numbers from immediately before and immediately after the Convention - or immediately before and immediately after Kerry's speech - you basically only have two times two sets left:

ABC/WaPo, bounce between the weekend before and the weekend after the Convention: +5-8%.
Rasmussen tracking poll, bounce between the day before and the day after the Convention: +2%

Newsweek, bounce between the two days of polling it did (the day before and the day after Kerry's speech): +8%
(that's a tricky indicator to use tho, cause one's basically halving the sample if one does)
Rasmussen tracking poll, bounce between the day before and the day after Kerry's speech: 0%.

All the other polls - Newsweek, CBS, etc - measure the bounce by comparing the numbers from (a span of days between) 1-3 days after the Convention, with numbers from 10 days, two weeks, even three weeks earlier.

But then again, thats how its always been done ...

Did I write "Friedman"? Poor Milton, I, of course meant, Krugman - what a face! - those thyroid-deficient, insolent, bulging eyes!

Kerry's bounce, whatever it may be, does require some squinting and torturing of the numbers to be seen. Hardly an enthusiastic precursor to an upset of an incumbent President.

I doubt that Bush will get much of a post -convention bounce either. My impression is that the American electorate is in a bit of a sullen mood. I believe, given their basic conservative bent (and by that I mean resistant to rapid change), and the uncertainty in the world, that Bush will win.

continued here

An important reminder:

We don't have an election in November; we have 51.

And each one is winner-take-all.

Here's some tasty goodness:

Pennsylvania
Survey USA (PDF). 7/31-8/2. MoE 3.7%. (6/10 results)

Kerry 53 (47)
Bush 41 (46)

Tennessee
Survey USA (PDF). 7/31-8/2. MoE 4.2%. (6/22 results)

Bush 48 (51)
Kerry 46 (41)

Washington
Survey USA (PDF). 7/31-8/2. MoE 4.2%. (6/4 results)

Kerry 51 (49)
Bush 43 (44)

Good news all around. One might even say, solid "convention bounces" for Kerry in these battleground states. (Tennessee a battleground state? Whooda thunk? )

Whatever else may be said, it appears nothing The Kerry Campaign has said so far has changed any minds.

Wow, 2% in Tennessee? Whaddaya know. I wasnt buying it from Zogby Interactive, but if another poll says so too ...

Edited overview of bounces to include Marist's non-bounce ... brings the average to something in between one and a half and two and a half percent (depending on whether you count the overnight Newsweek one).

I haven't run the calculations myself, but I heard on a radio discussion just a bit ago that if Zogby Interactive is excluded and if both the high-side and low-side single outliers are likewise excluded (a perfectly common statistical analysis practice), an apparent slight, though still less-than-margin-of-error, thus statistically insignificant , negative bounce occurs. One of the discussion participants posed the interesting, if essentially unprovable, hypothesis that indeed there had been a significant, and typical, post-convention Kerry bounce, but that the effect of same had been negated by an essentially unrelated-to-the-convention nearly equal-but-opposite contemporaneous shift of voter sentiment . The immediate, if in some cases transitory, post-convention polling gains, even if not in every case distinct leads or lead-changes, recorded by Carter twice, Ford, Dukakis, Mondale, Reagan twice, Bush the Elder twice, and Clinton twice, along with most recently Gore and Bush the Younger, all were referenced as supporting the hypothesis. Valid or not, a thought-provoking line of conjecture.

I've never heard of a national Zogby Interactive poll, I think those are only held state-by-state, in the battleground states. The national Zogby polls (for example the one released during the convention) are done by conventional standards and are as trustworthy as any other.

So any discussion about a bounce that talked about excluding Zogby Interactive numbers can at most (if it wasnt simply people talking out of their ass ...) be about the electoral college numbers. But measuring a bounce in terms of electoral college numbers seems extremely problematic to me, considering the wildly varying time intervals in which polls appear from state to state.


You mean the same point we've been making here, basically - but just in more difficult words? :wink:
->

I have a bone to pick with this article. I suspect some spin. I got suspicious at just about the first point Saletan chooses to make:


So - after mentioning three major polls and "the important data" in them that the media reports have missed, Saletan focuses us on one indicator we should look at to find out "what's changed": favourability. He then charges ahead with what the results from one of those three polls have on the matter. They clearly show us: the Convention greatly lifted Kerry's numbers.

Makes you immediately curious what the other two polls he mentioned registered on the question, no?

Well, in the CBS poll, Kerry's favourable rating, before the Convention, was 3% higher than his unfavourable rating. After the convention, it was all of 6% higher. His unfavourable numbers stayed the same; the favourable ones went up by all of 3% - just within the margin of error.

CNN? Again, post-Convention, Kerry's unfavourable stayed the same; his favourables went up by all of 3%. Within the margin of error.

"What's changed", the article asks - then proceeds to cherrypick an amazingly positive number thats also amazingly dissimilar from what other pollsters have on the same question.

Saletan proceeds to list many other numbers, many of which might well be spot-on. But when it starts out with a point like this, I can't help being a little sceptical.

Yeah, nimh, I think you pretty much called that one on-the-money earlier. I just thought you might find the anecdote interesting. I dunno exactly how they were considering what Zogby Poll, or all-in-all precisely which other polls they were discussing. And, as predictably common among the commentariat, I'm sure there was a more than just a bit bit of verbal flatulence involved ... there usually is when opinionators analyze anything

Someone once told us commentarians that we had nothing to lose but our chains ... ever since, the flatulence abounds in freedom. Shows ya how dangerous them revolutionaries be ...

Aux Barricades!

Heh ;-). And a propos of nothing, Houston seems an odd place for a lounge bar called The Proletariat ...