The Talk of the Town
Marc Ambinder reports:
Folks at my company are super excited about our newest product — a 300-sample-per-night daily tracking poll sponsored by the folks at Diageo. Last night’s track has Obama up six, 46 to 40 among registered voters.
In the past I’ve noted that the Gallup Daily Tracking Poll’s information is presented in a way that’s likely to generate buzz via misleading narratives — statistical noise looks meaningful the way they present it. But Hotline/Diageo are doing one better by using a much smaller sample, 300 people per day instead of 1000 per day. One bonus of the smaller sample is that it’ll be cheaper. And, ironically, another bonus of the smaller sample is that it’ll produce worse data — more random noise — that, in turn, will probably help gin up more interest in the poll.
19 Responses to “The Talk of the Town”
Register
RSSLogin
Topic Cloud
ArchivesBlog RollAbout Matt YglesiasContact Me
Donate
September 5th, 2008 at 7:56 pm
How wonderful… for Ambinder’s brand of journalism.
September 5th, 2008 at 8:05 pm
Diageo? The liquor company?
GoldenPalace.com should do a political poll, that would be another innovative way to get publicity.
September 5th, 2008 at 8:06 pm
Who stole Ambinder’s brain?
September 5th, 2008 at 9:14 pm
Thinkprogress Matt > Atlantic Matt
September 5th, 2008 at 9:51 pm
Look, it’s been clear for a long time that polls are just glorified man-on-the-street interviews for most media organizations. If they were really serious about it from a news standpoint, they’d all be reporting the pollster.com aggregation, or some suitable equivalent. That this is never done shows how fundamentally unserious the media is about polls.
[And also how little math they really understand. No surprise there.]
September 5th, 2008 at 10:08 pm
The error is related to the square root of the number of datapoints, so the error will only be roughly twice as bad, but the data will be three times cheaper. So …
September 5th, 2008 at 10:16 pm
I agree with the media aspect, but since everyone who actually cares uses poll aggregators now, this should give us better data overall — if weighted properly. No?
September 5th, 2008 at 11:31 pm
The commenter in me wants to say “is there a bigger dumbass/hack than Armbinder?” But the realist in me knows, yes, there are much bigger dumbasses/hacks than Armbinder, and that’s just sad.
September 5th, 2008 at 11:52 pm
Depends on what you mean by “bigger,” Seitz.
September 5th, 2008 at 11:59 pm
My new startup that I’m just about to found will do Diageo one better. We’ll do an hourly poll with a sample size of one. We’ll then weight that data according to a pseudorandom continuous variable somewhere between 0 and 100. So, for example, if my respondents opt for Obama 1 to 0, and the pseudorandom coefficient is 12.7, then that means McCain leads 87.3 – 12.7.
I expect that all the fluctuations in the poll will help draw attention to it. Just think of the headlines! “Obama threatens landslide”. “McCain makes miraculous 80-point recovery”. As long as you can convince yourself my numbers are reality, it’ll be great.
September 6th, 2008 at 12:05 am
A 900 sample over three days is not at all bad for a national poll (remember, you *want* to spread the poll across multiple days). The problem with tracking polls–the appearance of “fake” movement–is a somewhat separate problem, compounded by volatility endemic to this stage of the campaign.
September 6th, 2008 at 9:07 am
You will notice that Ambinder said that “folks” at his company were excited, not that he was excited.
September 6th, 2008 at 9:38 am
This week Diane Rehm had a major pollster on who admitted that they weren’t doing cell phones. So, all you youngsters without land lines get out and vote in November!
March 2nd, 2009 at 8:16 am
levitraExcellent site. It was pleasant to me.
March 22nd, 2009 at 7:15 am
tramadol
I want to say – thank you for this!
March 22nd, 2009 at 11:15 am
buy viagra online
Great site. Good info