The composition of the debate-watching audience is somewhat older and whiter than the electorate as a whole. At least, it was in 2008.
So, when CNN publishes a poll of debate watchers, and it's full of old white people - that's partly because that's who watches debates. It's also partly because it's a small poll, with N=430, so there will be a larger amount of error. And finally, you have to look carefully at who was polled:
Survey respondents were first interviewed as part of a random national sample on September 28-October 2, 2012. In those interviews, respondents indicated they planned to watch tonight's debate and were willing to be re-interviewed after theIt's not just a survey of those who watched the debate. It's those who planned ahead of time to watch the debate AND agreed to be re-interviewed. Could that select for a whiter and older demographic this year? Possibly.
Here's the numbers. CNN numbers are rough estimates from their published margins of error.
2012 CNN Poll 2008 Nielsen Numbers 2008 exit polls
67% Age 50+ 64% Age 45+ 53% Age 45+
67% Attend College 66% Attend College 76% Attend College
82% White 78% White 74% White
Note also that 81% of the 2008 VP debate audience was white.
Update: 2008 Nielsen Numbers above include those under 18 and, of course, those who are not registered to vote, so it's not an apples to apples comparison. 67% of those over 18 were over age 45.
And one last demographic: the CNN poll found ~30% of the debate watchers who are registered voters living in the South. Our own DailyKos poll, aggregated over the entire spring and summer seasons, has found 34% of registered voters in the South.
I think the reasons listed above can plausibly explain the CNN demographics.
So, this poll is not representative of the electorate as a whole (which CNN never claimed) nor is it likely precisely representative of the debate-watching portion of the electorate. It does, however, have a fairly reasonable demographic composition given the way in which it was conducted. That composition leans towards Republicans more than the electorate as a whole, as confirmed by Romney's above-water pre-debate favorability numbers (54% favorable).
Of course, Obama still lost according to the respondents no matter how you look at it. When 35% of Democrats say Romney did the best job, that tells you something. And yet - 45% said Obama was more likeable during the debate, 46% Romney. So Romney may have 'won,' but a large portion of those who thought he 'won' didn't like the way he did it.