It's actually pretty funny. And by funny I mean fraudulent.
Check out how the numbers break down.
Here is how today's CBS poll opens:
"(CBS) The presidential race between Barack Obama and John McCain is now even at 42 percent, according to a new CBS News poll conducted Monday-Wednesday of this week. Twelve percent are undecided according to the poll, and one percent said they wouldn't vote.
This is in contrast to a poll conducted last weekend, where the Obama-Biden ticket led McCain-Palin by eight points, 48 percent to 40 percent." (boldface mine)
I guess they didn't count on anyone actually reading both polls to the end and doing some quick 'rithmetic.
Here's the nitty-gritty on the last two CBS polls.
I'll start small.
August 29 - 31, 2008 (Total Respondents - 875)
UNWEIGHTED WEIGHTED
Total Registered Voters 781 743
Total Republicans 247 229
Total Democrats 310 308
Total Independents 318 337
Oddly enough, the "weighted" totals of 229 (R) + 308 (D) + 337 (I) add up to the TOTAL number of respondents... minus one.
That's right:
Unweighted = 875
Weighted = 874
I guess whoever conducted the survey decided not to count their crazy uncle Willamina after all.
It gets better.
September 1-3, 2008 (Total Respondents - 835)
UNWEIGHTED WEIGHTED
Total Registered Voters 734 691
Total Republicans 229 215
Total Democrats 250 241
Total Independents 255 235
And here are the questions they asked:
VOTE FOR PRESIDENT (Among registered voters) (Hence the 42%-42%)
DESCRIBE YOUR SUPPORT (Among registered voters)
OPINIONS OF THE CANDIDATES (Among registered voters)
DOES HE UNDERSTAND THE NEEDS AND PROBLEMS OF PEOPLE LIKE YOU? (Among registered voters)
...hang on a minute. ALL of these responses are Registered Voters only.
UNWEIGHTED WEIGHTED
Total Registered Voters 734 691
But wait a sec, I thought they surveyed 835 people, not 734. What happened to everything those other 101 people said?
...oh, who knows.
All right, enough build up. Here's where the actual WTF-ery begins.
In the Aug 29-31 poll, where Obama came out 50-42, the ratio of weighted Republicans surveyed to weighted Democrats surveyed was 30.8% (R) to 41.4% (D).
In the Sep 1-3 poll, where he and McCain tied at 42, the same ratio is 31.1% to 34.8%.
If you increase the percentage of Republicans in your poll by 5% of the total, guess what? Odds are that close to 5% more of all of your respondents will favor the Republican candidate!
For the record, Rasmussen uses +5.7% in favor of Dems in his surveys, and Gallup uses +9% in favor of Dems.
Why?
Because there are a whole lot more Democrats out there than Republicans this time around!
(They also tend to use between 2,500 and 3,000 respondents instead of ~700-800. Because they apparently understand that when you use that few people nationally, real data begins to be drowned out by statistical noise.)
Now, I understand the media wants a suspenseful narrative and all that nonsense, but come on now, CBS. You have to try a whee bit harder than that to make the numbers lie for you. Simply comparing the results of polls taken with completely different Party ID methodologies and hoping no one notices?
FAIL.