Articles and Analysis


"Internal" Measures of Change on the CBS Survey

Topics: 2008 , Barack Obama , Bounce , CBS , John McCain

The just released CBS survey we posted a few minutes ago is one of the first of the post Democratic convention polls to include extensive "internal" measures on perceptions of the candidates. If you want to get beyond the ongoing obsession with convention bounces, the CBS results pdf is worth reading in full.

Most useful within in are the measures that show the biggest changes since the pre-convention CBS poll conducted in mid August. Some examples:

  • 50% say Obama "has made it clear what he would do as president;" up 9 points (from 41%)
  • 58% say Obama is "tough enough;" up 10 points (from 48% in July)
  • 67% of Obama voters say they "enthusiastically support" him; up 19 points (from 48%) before the convention
  • 37% rate Joe Biden favorably; up 23 points (from 13%)
  • 41% rate Michelle Obama favorably; up 13 points (from 28%)
  • 52% rate Hillary Clinton favorably; up 7 points (from 45%)

At the same time, several measures show little or no change since mid-August:

  • 39% rate Obama favorably; unchanged since mid-August
  • 44% say Obama has "prepared himself well enough for the job of president;" unchanged since mid-August
  • McCain's ratings are unchanged; from 34-33% favorable/unfavorable in mid-August to 35-34% now.
  • 48% say McCain "will generally continue George W. Bush's policies;" up only one point (from 47%). 

Tune in again next week to see how the Republican convention alters perceptions of McCain and Obama.



Don't go grocery shopping with someone from CNN, because they do not know their apples from their oranges.

I was puzzled about this statement in today's "CNN poll-of-polls":

“Any bounce Obama might have received out of his party’s convention appears to have largely negated by McCain’s surprise selection of Gov. Palin as his running mate,” said CNN Political Editor Mark Preston. “This race has been close all along, and is unlikely to change.”

What got me scratching my head is that today's CNN poll-of-polls includes CNN/ORC, USAToday/Gallup, and CBS, the latter two are reporting post-convention bounces for Obama in their poll analysis and even the CNN/ORC poll shows a 1 point movement towards Obama (this poll increasingly appears to be an outlier, btw).

Mystery solved when I looked at the previous CNN poll-of-polls that serves as the basis for comparison. (cut and paste this into your browser, since my posting would otherwise have too many URLs in it and be held up for approval):


In this poll-of-polls, the average came from ABC/WashPost, Fox/Opinion Dynamics, and the Gallup Daily tracker. I do not believe the two other organizations have released new polls yet, but the Gallup tracker shows a significant Obama bump over the course of the convention, too.

Bottom line, CNN is cherry picking polls to further the finding in the CNN/ORC poll that there is no bounce for Obama. This is truely terrible, dishonest reporting. What's dumb is that we can expect a bounce for McCain, too, which most likely will cancel out the Obama bounce. But for CNN, we will go through two conventions and they will report that the conventions had no effect when, in fact, some very real changes in public opinion have occurred - all so that they can protect their precious black-box CNN/ORC poll from criticism.


Looking back at my posting, I should add that all the polls in the CNN poll-of-polls were conducted after the Palin announcement, too. Palin did not cancel out an Obama convention bounce, as is alleged by CNN.


Mark Lindeman:

Michael, I'm confused: are Team CNN actually "cherry picking," or are they just being totally indiscriminate about the polls they compare? or can't we tell? I don't know that I'm doing Team CNN any favor by raising this question.

Regardless, the conclusion that the Palin announcement "largely negated" any Obama bounce is indeed weird. There might be a decent case that Obama's bounce was on the small side -- which, intuitively, would not be startling considering how much exposure he has received already this year. Whatever. I'm more interested in where the race will stand next Tuesday or so.

In the spirit of Mark B.'s post, I will offer one (superficial) comment on the internals: there sure are a lot of people who pronounce themselves "undecided" (not favorable or unfavorable) about Obama and McCain, given the explicit choice. No big surprise, but always good to keep in mind.


@Mark: fair enough. I'm willing to agree that CNN is just being stupid rather than self-serving. Either way, they should know better.



If half the country thinks George - I mean, John McCain is going to continue Bush's policies, no wonder he had to shake things up... But I am a little surprised that number didn't move more despite the Democratic Convention. Maybe all the ads Obama's been running so far have been effective enough... The only positive I can see is that 20% now think McCain's policies would be "less conservative" than Bush's, down from 24% pre-DNC (maybe within MOE, but hey...)

To Michael/MarkL/MarkB:
Wouldn't it be easier and better for CNN to just take the numbers from Pollster.com's regularly updated horse race numbers?
At the same time, would it be possible to modify the current trend plots so one can read the values at different dates along the lines? Or at least, more grid-lines... Thanks.



I don't know, I was going to agree with your first conclusion, Michael, that CNN are being self-serving, as there must surely be some bottom line of compentence, but then I read their last line, "The poll of polls does not have a margin of error". Not that they couldn't calculate it but the poll of polls has no possibility of sampling error. Hence I am now tending to stupidity.


Mark Lindeman:

@RS: Yes, it would be better for CNN to use pollster.com. If they're reluctant to do that, then they could (e.g.) consistently use the same polls. Crazy talk, I'm sure.

@John: To say that the poll of polls "does not have a margin of error" isn't to say that it is error-free. It's to say that one can't calculate a margin of error -- presumably because the surveys aren't interchangeable, so one can't simply add the sample sizes together and calculate a new, smaller MoE. The error isn't in saying that; the error is in ignoring it, to the extent of making indiscriminate comparisons. (The pollster.com average is also pretty indiscriminate -- omnivorous, even -- but should be more robust because it doesn't use an arbitrary last-three rule.)



I have been of the opinion for sometime that both Fox and CNN are in the News Entertainment industry.

In CNN's case their own polls do give you anything other than headline numbers that they want to give.

The poll of polls nuff said.

With regards this CBS poll, the information, it's pretty well bang on to what you would have expected the Dems convention to deliver.

The Convention was the best organized and disciplined probably of all time.



Mark, of course I realise that it is extremely difficult to calculate a MOE from multiple polls with different pollsters, but at the end of the day the poll of polls is still a sample and therefore must have sampling error. CNN stated that the poll of polls does not have a margin of error. This is factually incorrect. If one cannot calculate parameter A of object B, it is incorrect to say Object B does not have parameter A.


Mark Lindeman:

@John: Rather than debate whether a margin of error that can't be calculated nonetheless exists, it would probably be cleaner to argue that it can be calculated meta-analytically, although people would debate the assumptions. Regardless, your original post attributed to CNN the view that "the poll of polls has no possibility of sampling error" (quite distinct from whether its MoE can be calculated). I don't think that is a fair reading.

Of course, Mark Preston seems to be talking about the issue as if sampling error were zero. So maybe it doesn't matter what the sentence we're discussing was supposed to mean.



Sorry, Mark, I see what you are saying now, that sampling error goes beyond the margin of error, which is of course right. I should of been more precise in my orginal post, although I still maintain their statement was incorrect. But as you say, given the way Mark Preston was talking about issue, this debate is a bit academic.


Post a comment

Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.