Wednesday, October 16, 2013

There's a winner every time

As a general rule, if you take newspaper headlines about survey results at face value, you should stay away from those friendly games of chance at the county fair. In this case, though, WashTimes readers have a little extra reason to be peeved, because the home team is lying to them.

OK, maybe not technically lying. Unlike "rising" or "falling," which are negatable, "in free fall" is just impressionistic enough that the journalistic expert involved can claim, in effect: Well, it looks like free fall to me! And in an abstract world, you'd kind of like the idea of an expert who defies the apparent evidence to tell you what's really going on. The trouble with this one is that there are no meaningful values of "free fall" -- and almost none of "fall" -- for which the claim is true.

Democrat Cory Booker’s lead over Republican Steve Lonegan in the race to fill New Jersey’s vacant Senate seat has shrunk to 10 points with about 48 hours to go before Wednesday’s special election.

No doubt it's pretty to think so, and it'd be cherry-picking to point out that in the most recent poll, the Democrat's support had fallen from 52 percent to 54 percent. (Let's just use the polls collected at the generally execrable RealClearPolitics, for the heck of it.) No doubt the Times is extrapolating from the poll before that, in which he fell from 58 to 52 percent. But that ignores the previous poll, in which he fell from 53 to 58, or the one before that, in which he fell from 50 to 53 -- starting to get the idea here?

Now, to be fair, the Times is talking about the evil communist's "lead" -- which before expanding to 14 points shrank to 10 points before expanding to 22 points before expanding to 12 points before shrinking to 11 points ... OK, by now you should be agreeing that the stupidest thing you can do with quantitative data (aside from making it up, which is evil rather than stupid) is to draw conclusions from the numbers you like, rather than the numbers you see.

There is, of course, no such thing as an "average" of the random stuff that shows up as poll results. That doesn't mean you can't draw conclusions from random collections of polls. You could, for example, look at the RCP data and wonder if there's an association between small samples and really good results for Booker (sure looks like it). You could also conclude that unknown candidates tend to gain support as they become better known. That's one well-known way in which a lead can "shrink" without the leading candidates' support having gone into the mythical "free fall."

You can do a lot, in general, with numbers. If you want to be taken seriously when you talk about them, though, you should avoid making stuff up. That's a particular challenge for the WashTimes, for which making stuff up is an important part of the workday, but in this case it's especially unfortunate. Times readers might, after all, wonder why they're being told the world is all rosy when indeed it's falling in on them. And when the preferred audience is stirring the tar and stockpiling the feathers, you really might want to think twice about reminding them that there's a winner every time.

Labels: ,


Blogger Strayhorn said...

I've said it before - polls belong on the same page as the horoscopes.

9:16 AM, October 17, 2013  
Blogger The Ridger, FCD said...

On the other hand, the Times did feature a guy in the 21st Century NRA running feature - a nice wrap-around with color portrait of this week's pinup, on top of the front page. Until this week I'd thought everyone in the 21st Century NRA (might be different from the regular old one, I suppose) was a pretty young(ish) white woman.

At least they're still all white.

6:53 PM, October 17, 2013  
Blogger The Ridger, FCD said...

Oops. That's "3rd Century NRA"

7:07 PM, October 17, 2013  

Post a Comment

Links to this post:

Create a Link

<< Home