Diageo/Hotline: Has the race tightened, or are D/H cooking the figures...?

[regnaD kciN]

Check it out, because you're going to be hearing news that "Obama's lead has shrunk dramatically over the past day" in this poll, going from +6 to +2 in one day.

There's only one problem with these results:

Yesterday, D/H showed Obama ahead 47%-41%, with a party breakdown of 41%D - 36%R - 19%I.
Today, it's only 46%-44%, with a party breakdown of 40%D - 38%R - 18%I.

That's right -- they changed the weighting from a 5% advantage to merely a 2% advantage in favor of Democrats.

:wtf:

If you go back to the original weighting, you'd find today's results would still be around the same +6 for Obama as in yesterday's results. But I'd guess that wouldn't get the Corporate Media's attention...especially when they need a "McCain's catching up quickly!" meme to promote tonight's debate.

By the way, this is not the first time this has happened. When D/H first came out, I thought its numbers were a bit too optimistic for Obama. As it turned out, they dropped suddenly about three weeks ago, too. Sure enough, it was due to the same reason. They started out with a 9% partisan advantage for Democrats, and had dropped it to 5%. Now it's suddenly 2%. Political whoring, anyone?

(I don't deny that, if Democrats only outnumber Republicans by 2% at the polls on Election Day, it's going to be closer than the mainstream polls indicate. But, as far as I'm concerned, if we can't do any better than a 2% edge on November 4th, we deserve to lose. I figure at least a 5%-6% edge, and probably more.)

[Drunken Irishman]

1. Not cooking their numbers. D/H just has a higher statistical noise than the other daily trackers.

[jefferson_dem]

2. Thanks for the analysis.

I thought those partisan numbers were not right.

Are we sure they weighted the data to 46/44 or are they just reporting the partisan composition of the sample - and just happened to connect with more Repugs than Dems yesterday?

[regnaD kciN]

6. Well, since they don't report internals...

...and the party breakdown for the entire three-day sample shifted from 41/36/19 to 40/38/18 overnight, either they're re-weighting or they somehow only managed to reach Republicans last night. :eyes:

But, if it was just statistical noise, you'd expect the party breakdown to change virtually every day, as old days get tossed out and new ones entered. But, according to the commentary I've seen so far elsewhere, there have only been two shifts in party breakdown since the poll started just after the conventions: one from +9D to +5D, and now one from +5D to +2D.

[MadBadger]

3. They arent cooking their numbers, just a bad day of polling.

[regnaD kciN]

8. See my comment above...

I see no way a single "bad day of polling" could skew the party ID of a three-day tracker so drastically, especially when it hasn't been shifting daily before.

[Jersey Devil]

4. Also consider it is a 1 day poll and not a rolling average like Ras and Gallup

so it is more prone to volatility even without messing with the party breakdowns.

[regnaD kciN]

7. Not true...

...as I understand it, D/H is a three-day tracker just like the others.

[maseman]

5. I work in research

Usually you would nto change methodology during a study. In other words if after the elections are over they deem that the Dem-Rep-Ind levels were off they would "fix" it for the next election cycle. But to change this kind of methodology during the polling so close to the election is nuts UNLESS you CLEARLY post what the changes are and why the data changed.

[uponit7771]

9. Right!!! Their method should stay fairly static...I'm discounting them some because of this change