In How to Slant a Poll 101, John Hawkins at Right Wing News explains how CBS gets its headline results:
The first thing you have to understand is that there are 3 different groups of voters the media may poll: adults, registered voters, and likely voters. Out of these 3 categories, “likely voters” is the group that almost always turns out to be closest to the actual election results while “adults” is the group that slants most heavily towards Democrats. Although it’s difficult to pinpoint exactly how much polling “adults” instead of “likely voters” slants the poll results to the left (when compared to election results), it’s probably somewhere between 5-10 points. So, let’s split the difference and say 7.5 points.
So, it seems likely that Bush’s approval would probably be somewhere around 41.5% if this had been a poll of likely voters. Still, pretty bad.
But, there’s another factor we haven’t adjusted for: the percentage of Republicans, Democrats, and Independents who participated in the poll. In the 2004 election, the breakdown by party was as follows:
While those numbers can change and do change over time and there’s no set rule that says for a poll to be fair those percentages should match up exactly with the breakdown from the last election, the numbers should be pretty close.
So, let’s look at the weighted party breakdown from the CBS poll: 1018
Democrats: 37.4% (381)
Republicans: 28.4% (289)
Independents: 34.2% (348)
So, they undersampled the number of Republicans by more than 8.5% and over sampled Independents by more than 8%. Let’s adjust for that (in a very general way). Add in 8 more Republicans and we’ll say Bush’s favorability goes up 8 points. Take out 8 Independents and we’ll figure Bush loses 4 points of support (Independents were roughly split between Bush and Kerry in 2004) and now Bush’s approval rating, after having 4 points added onto it, is at 45.5. Of course, it’s not quite as simple as I’ve made it look here, nor is this as accurate as simply polling likely voters with a correct breakdown of party affiliation, but it’s close enough for our purposes.
Then, we consider the polls margin of error, 3 points, and Bush’s real approval rating among voters who’ll actually be going to the polls in November is probably somewhere roughly between 42.5 – 48.5. That’s not great, but it doesn’t have exactly the same sort of zing that 34% has either, does it?
This isn’t news. It’s just more political partisanship from the most dishonest network.
Hat tip to Memeorandum.