I don’t know why I watch or read CNN anymore. I saw this story on the President’s speech and read the following:
In the post-speech sample, 70 percent of respondents said Bush’s policies on health care were positive, while 66 percent approved of the president’s plan for Social Security.
Bush showed almost as much improvement on Iraq, with 78 percent of respondents saying U.S. policy there is heading in the right direction, a 12 percentage point increase over pre-speech polling. Overall, 77 percent of respondents said Bush is taking the country in the right direction after the speech compared to 67 percent beforehand.
I thought that sounded unbelievable. Turns out it is, because in the next paragraph you read this:
The strong positives for the president’s policies may in part be a reflection of the poll’s sample. Of the 485 people surveyed, 52 percent identified themselves as Republicans, 25 percent as Democrats and 22 percent as independents. The poll was done by telephone interviews and has a margin of error of plus or minus 5 percentage points.
Ya think? Ya think that if you double the number of Republicans, it might skew the poll a touch?
Idiots.
rachelrachel
SOTU polls always get skewed in favor of the president, because his supporters are more likely to watch than his detractors.
This occurs regardless of the party affiliation of the president.
The sited article is better than some I’ve seen, because it at least explains the discrepancy.
Sam
Who in their right mind (no pun) would want to identify himself as a Democrat after the ideological pummeling they’ve taken this week? Conversely, who WOULDN’T want to be ID’ed a GOPer this week?
I’m surprised that 25% found the guts to group themselves with Teddy Glug-Glug, Nancy P, and Dan Rather.
I’m guessing those 22% “Indies” were largely Dems before November 2004…
big dirigible
Interesting that they use the same old error formula, square root of N over N, as always, to get their error figure. That’s hard to reconcile with the fact that the sample is so obviously skewed. The problem is that the error formula is based on Gaussian variables, and doesn’t account for systematic error. Discussion of statistical errors has been convered on some blogs, including mine, but the subject is terminally boring.