Friday, December 04, 2009

On clues and having them

I think the cousins at McClatchy (along with Charlotte and any other MCT fishwraps or subscribers that ran this lede) owe the reading public a correction or two:

With the jobless rate rising and his approval ratings sinking, President Barack Obama hosted academics and leaders of business and labor at a White House jobs summit Thursday, seeking advice on how to boost employment.

If you're the sort of paper that promises to correct errors of fact, it's time to get to work. First up is that rising unemployment rate -- in a story written, as noted in the eighth graf, "a day before the Labor Department releases November unemployment statistics." In a way, it's bad luck for the writers that the November stats show "an unexpected dip" in unemployment, but it's bad luck enabled by bad judgment. We have a really simple rule of basic journalism for that: Write about what happened, not about what you think might happen. That way, you don't have to write corrections when your guess turns out to be wrong.*

You can see why the lede looked attractive, though; that concise pairing of contrasting elements just captures the old zeitgeist perfectly, doesn't it? Trouble is, the second element is no truer than the first -- it's not as demonstrably false, but it can only be true with a reading of the available evidence that's so selective as to be deliberately distorted.

Obama's decline in the ratings is the sort of conventional wisdom that drives much of the discourse at Fox, where it's popular because it makes sense and helps other things make sense. In the reality-based world, we're supposed to rely on evidence -- that being, after all, where journalism gets its claim to be objective. So let's look at the roundup of polls over at RealClearPolitics (not the "RCP Average," which is a meaningless number no responsible journalist would ever allow into print).

Obama's job approval in the Gallup poll, with an N of around 1,500 adults, is a point higher than it was last month (52% vs. 51%). Rasmussen (N = 1,500 likelies) is unchanged at 46%. USAT/Gallup, with around 1,000 adults, is unchanged from October at 50% (bigger N, for some unknown reason). Quinnipiac's big-N survey (2,000-plus registered voters) is at 48% for early-ish November, down from 50% at the end of September. The brand-new CNN poll (48%, 1,041 adults) looks like a significant decline from two weeks ago, 55%, but we want to look at that a little closer. It could mean any of several things:

1) The change in the sample value represents a real change in the population value
2) The three previous CNN polls (55%, 54%, 55%) closely approximated the population value and this one is a far outlier -- the one case in 20 (at 95% confidence) in which the sample lies outside the confidence interval, or "margin of sampling error"
3) The real population value is around 51%, so all four samples are accurate (if extreme) nonchance representations of the population

What's the safe way to put it? Were I in a mood to generalize, I'd say that the president's job approval declined significantly from summer into fall but appears to have kinda-sorta leveled out. At Gallup, it's a point higher than in mid-September; at Rasmussen, it's a point lower. (Yes, survey data is a lot less sexy when you do it right.) And that makes our lede wrong on the second count as well. You can believe whatever you want to about public opinion,** but if you want to proclaim facts and attribute them to evidence, sorry -- you don't get to make things up.

Does that make McClatchy biased? Yes, but not in the partisan way that tends to be associated with the "media bias" that people like to yammer about. It's biased toward explaining things in terms of conventional wisdom, without regard toward whether the conventional wisdom is either true or informative. That's probably less offensive than the open and deliberate biases on offer at the Fair 'n' Balanced Network (and it certainly isn't as pervasive), but in terms of adding to or subtracting from the sum of human knowledge, it's hard to see how it's any better.

* A lede that said unemployment was expected to have been found to have risen again in the upcoming monthly report would have been accurate.
** If any of the MCT folks want to argue about confidence levels, the address is at right.

Labels: ,


Post a Comment

Links to this post:

Create a Link

<< Home