Sunday, September 21, 2008

Shiny jangly stuff

The craft of design is generally held in high regard around here. That's "design," the use of visual principles to help interpret and organize the day's events, not "design" as in "Let's throw some shiny jangly stuff at the page and hope people with the attention span of garden slugs will put some quarters in the rack and be temporarily awed." Guess which approach is in charge at Fort Lauderdale?

Once you've figured out from the new! improved! nameplate which paper you're reading ("Look, honey, Superman has his own newspaper!"), you might first notice that there isn't a lot of "news" out there in the world. Then you might notice that the most Cosmically Super-Important Thing Going is a public opinion survey. It looks like about two-thirds of the page; is that space well spent? There are a couple of ways to look at that: how the space is spent and what it's spent on.

The top hed -- the "How Florida would vote" thing floating around in the air above the dominant visual element -- is a good place to start. At best, it's overawed by the obvious: "Who would you vote for today?" is what presidential preference polls ask to determine -- you know, presidential preference. But it gets worse in a hurry. By turning the individual preference question into an assertion about what the state "would" do, it skids into nonsense. The statement can't be true; if nothing else, there wouldn't be an "undecided" (the 6% under Obama's left ear) if the election "were held today."

The visual centerpiece is striking for how little it conveys about anything. In case you didn't notice, it's a couple of mug shots, a couple of numbers, and some stylized parts of what looks like it was supposed to be an American flag. Thus, the entirety of "content" above the fold at the Sun-Sentinel can be summed up in this sentence:

"A survey of 600 likely voters in Florida, conducted Sept. 15-18, found that 46% said they would vote for McCain and 45% for Obama if the election were held today."

That's actually more information than the page gives you, since it includes the sample size and the dates the poll was in the field. Without a lot of effort, we could also add the maximum margin of sampling error (and the all-important confidence level), reminding us that the designers aren't the only ones at fault here.

Is this starting to sound like a broken record? There is no such thing as a "statistical tie" (just as your stylebook says, if you follow AP style). "Within the margin of error" is a meaningless term.* So a graf like this:

Obama has overcome McCain's earlier advantage in the state to reach a statistical tie of 45 percent to 46 percent, according to a statewide poll of likely voters conducted last week for the Sun Sentinel and the Florida Times-Union. The 1-percentage-point difference is within the poll's margin of error.

... suggests a fundamental misunderstanding of the subject matter. It doesn't matter whether the New York Times gets it or not, or whether the story is from your Washington bureau chief. It's like reading that someone "scored a home run"; if you don't know how the scoring works, you undermine any conclusions you draw about the game.

There's a bigger, or at least a two-layered, problem as well. The story contends that Obama has overcome "McCain's earlier advantage," apparently based on this assertion: "Most pre-convention polls in Florida showed Obama trailing McCain beyond the margin of error." That suggests somebody isn't reading the paper: At midweek, the Sun-Sentinel was reporting a Time poll with the candidates at 48-48. At the beginning of the month, something that looks like an editorial proclaimed that "according to the latest polls," Florida was "pretty much up for grabs." Judging from the RCP log,*** a good way of stating the "most polls" thing would be: Some do. Some don't.

The best conclusion a careful newspaper would draw from all this is: Reply (still) hazy. Ask again later. This is a competently done poll that doesn't come anywhere near supporting the end-of-the-world treatment it got. Whoever's fault that was, please stop.

*If you're having trouble convincing a writer of this, feel free to call the 24-hour help line.**
** Or, until the help line is staffed, drop a note and we will try to help.
*** Good for results of individual surveys; just stay from the alleged "average."

Labels: ,


Post a Comment

Links to this post:

Create a Link

<< Home