Robert Waldmann tells us that Billmon is back:
Billmon (September 10): The Washington Post Lies About Its Own Poll: A few hours ago the Post put up a front page story reporting the results of the latest Washington Post-ABC News poll. And said story appears to be baldly misrepresenting -- and by misrepresenting I mean lying about -- the size of Obama's post-convention bounce.
Either that, or the story's authors are too freaking incompetent to read their own polls.
The story, by long-time political reporter Dan Balz and some guy I've never heard of before named Jon Cohen, says this:
The survey shows that the race remains close among likely voters, with Obama at 49 percent and Romney at 48 percent, virtually unchanged from a poll taken just before the convention.
And indeed, The last Post-ABC poll, released Aug. 27 (first day of the GOP convention) had the race at Obama 46 / Romney 47 -- a statistically insignificant difference from the latest poll.
There's just one problem: The previous head-to-head results (46 / 47) were among REGISTERED VOTERS, while the new ones (49 / 48) are for LIKELY VOTERS. Apples and oranges.
In reality (i.e. apples to apples), the move towards Obama in the Post-ABC poll has been fairly substantial: from 46 / 47 in Romney's favor in the last poll (again, among REGISTERED VOTERS), to a 50 / 44 lead for Obama now, a net swing of 7 points. Which isn't quite outside the MOE, but is still highly unlikely to be just random noise.
But almost all of that bounce was swallowed up by the Post-ABC likely voter screen…. [C]omparing registered voter samples to likely voter samples is one of those basic factual errors that cause me to want to grind my teeth (and grind up certain parts of the offending reporter's body)…. It's awfully hard to write it off as a "honest" error, given that we're at an extremely critical point in the race, and Team Romney is beginning to take heat from the GOP establishment because of its crappy poll numbers.
It's actually even weirder than Billmon states. Here are the first three paragraphs:
Last week’s Democratic National Convention helped President Obama improve his standing against Republican Mitt Romney, according to a new Washington Post-ABC News poll, but did little to reduce voter concern about his handling of the economy.
So now we expect a second paragraph about how Obama has improved vis-a-vis Romney, and perhaps information about continued discontent about the economy. But that's not what we get:
The survey shows that the race remains close among likely voters, with Obama at 49 percent and Romney at 48 percent, virtually unchanged from a poll taken just before the conventions.
And this is where Billmon's remarks about likely vs. registered voters apply: Balz and Cohen have made a big bounce disappear.
Then, however, Balz and Cohen reveal that they do understand the difference between likely and registered voters after all:
But among a wider sample of all registered voters, Obama holds an apparent edge, topping Romney at 50 percent to 44 percent, and has clear advantages on important issues in the campaign when compared with his rival.
But they don't say that the post-convention registered voter poll is very different from the pre-convention registered voter poll.
That's an awful lot of pretzel-like contortions to keep from writing the "Obama is getting a big convention bounce" story that the poll tells us...