Those astonishing polls on the debate are not what they seem to be

Andrew B. writes:

This is in response to your report that a CBS poll and a CNN poll on the debate showed Romney winning by 65-34 and 54-40 on the economy, and handliy defeating Obama on most other issues as well.

Both CNN and CBS noted their debate poll samples were more heavily Republican and had fewer Independents than their national opinion poll samples by about eight points. This is borne out by surveys from Google and Reuters which pegged an Obama win by a margin of 48-31, or about that same eight points higher than the TV poll margins.

It is to be expected that polling more Republicans will find more people who like Romney on the issues, just as doing the same with Democrats would make you think you were surrounded by Obama lovers.

What the polls really tell you is that more Republicans tuned in to watch last night than their share of the general population, and a large percentage of them took away an impression that their man lost even if they preferred his positions.

Looking at the ethnic breakdown of the audience at the event, it would not be surprising that many of Obama’s minority supporters did not bother to watch at home, leading to a more heavily Republican national TV audience. That doesn’t mean those same minority voters will not turn out in droves to vote on November 6. Even in 2008, despite multiple cycles of turnout growth, urban minority turnout was still far lower than white turnout. Republicans have to be aware of the importance of driving their voter turnout much higher than 2004 and 2008, probably on the order of 25 percent, to counteract Obama’s get-out-the-vote effort and give themselves a chance to win. They are going to need to hit something like an 80 percent or higher turnout of registered Republicans.

LA replies:

Thank you for this. Since I would assume that most polling organizations are reasonably professional most of the time, I would also assume that they ask respondents their party affiliation and make adjustments in their numbers so that the poll will reflect nationwide party affiliation. Obviously a poll in which a significant majority of the respondents are Republicans tells us nothing about the actual electorate and is deeply misleading if presented as a representative national sampling .

These problems in polls mean that one must research each poll before deciding to post it, which is not very practicable for a blogger.

Posted by Lawrence Auster at October 17, 2012 08:18 PM | Send