Thursday, November 08, 2007

There's one born every minute

Wait. Here's a better idea. How about we call you "happy trotting elves of the public relations craft"*?

Sorry. It's a bit label-like for a hed. But you asked. At least, the Tampa Tribune sort of did. So let's play along for a second, as if this had been some sort of real news, for therein lie some hints about how the little satans** of the PR industry manage to swap their handfuls of magic beans for our cattle.

Here's the lede. And before you single the reporter out for all the well-deserved ridicule, bear in mind the number of other people who played along: graphics, design, copydesk and originating eds.

Coffee! Soda! Chocolate! Call Us Caf-Fiends
TAMPA - The survey released Wednesday was clear: The Tampa-St. Petersburg-Clearwater area is the second most caffeinated metro area in the United States. Only Chicago ingests more.

The press release was clear enough. How clear was the "survey"?

Coffee, soft drinks, tea, chocolate, pain relievers, energy drinks, caffeine pills. We love it all, if HealthSaver discount health care service's survey of 2,035 people in the country's 20 largest urban areas is to be believed.

And there's a second side in play too. In addition to asking about the survey, we have to ask about the reporter's ability to read the results -- at least, the ones the polling company put into its PDF (it's a 67-page file, but don't worry; it goes by in a hurry). Unfortunately for him, the "survey" doesn't say anything of the "sort." Ninety percent or more of respondents report "less than once a week" or "never" for their use of energy drinks and caffeine-containing pain pills. For caffeine pills themselves, the "never" figure is 97%. Which, in a legitimate sample, is pretty conclusive as to whether "we" "love it all." That'd be a straight-up fabrication.

Since the reporter and his unnamed enablers evidently can't be bothered to pay attention to details, are they at least regurgitating legitimate facts? Well, sorta. In broad-brush terms, the survey technique uses a legit method, in that it randomly samples telephone numbers. The devil-in-the-details part is how and which, and then in how the responses are assembled into the bright shiny numbers that seem to make press releases so attractive.

The survey includes 2,035 people total, for which the press release reports a margin of sampling error of 2 percentage points. It's not nice to round confidence intervals; at 95% confidence, that margin is more like 2.2 points. And the press release doesn't include the margin for the 20 different metro areas. If you look at the full "report," you'll find -- surprise! -- a margin "for any given MSA" of 9 points (again, understated; taking 100 to stand for "about 100" as the N in each MSA, it's more like 9.8). Meaning -- and this will be on the final -- we're 95% confident that given a sample mean of 50, the population mean would fall somewhere between 40 and 60. Given that our population is "listed phone numbers," make of that what you will (hey, Tampa did).

Now, how about that "caffeine index" that has Tampa all a-twitch? That's an average of "uses" of each of the caffeine products on the list. All uses are weighted equally, we're told, so one venti equals one Milky Way equals one tea, and so forth. What happens next is a bit confusing. The questions themselves appear to represent "every day," 4-6 times a week" and so forth, though the methods page says values of "7 or more times a week" -- which isn't the same thing as "every day" -- are all set to 7. Then the 4-6 category is set to 4, the 1-3 category is set to 1, "less than once a week" is at 0.5, and "never" is 0.

See a problem there with turning ordinal data into ratio data? A person who has one Milky Way a month (0.5) and three caffeinated pain relievers a year (0.5) is the same as somebody who drinks coffee three times a week (1.0)? But somebody who drinks coffee four times a week counts four times as much as the person who drinks coffee three times a week?

Anyway, add 'em up, set the overall mean at 100, and index from there to show the "relative variance," so a city scoring 200 is twice as caffeinated as the average. So Tampa residents with listed phone numbers (give or take about 10%) average 14.5% more caffeine "uses" per week than people with listed phone numbers in the nation's 20 largest MSAs as a whole? Just checking.

In short, the survey can be "believed," as long as you bother to ask what it's saying and -- crucially -- don't mind that the variables don't represent anything of any practical significance or relevance. But that's evidently not the point:

HealthSaver did the survey, a spokesman said, to "stimulate awareness of caffeine consumption and to shine a light on the numerous health benefits of caffeine."

Was there a particular reason we believed that? Just wondering.

News wires buzzed Wednesday about the survey.

Not that I can tell. The thing that did the buzzing was the PR Newswire, which -- now let's not always see the same hands -- sends out press releases. (Although poor benighted UPI seems to have fallen for it too). And a couple of Business Journal pubs appear to have gone along. Otherwise, a quick and nonrandom check of frontpages and Web sites of papers in the other metro areas suggests that America's Dailies were asleep at the switch. Even in Miani (109.9), Mesa (106.1) and Houston (103.8).

The idea that Tampa would be a runner-up caffeine mecca was a revelation to morning TV anchor Diane Sawyer, who was taken aback at the news during Wednesday's broadcast of "Good Morning America." Many assumed Seattle, home to Starbucks, would be No. 1 on the list, but it only got to No. 9.

They certainly waste a lot of time on morning TV, don't they?

This information did not exactly surprise 69-year-old Danilo Fernandez.

Sitting Wednesday afternoon at a small table inside El Molino coffee shop in Ybor City, he had just emptied a foam cup full of espresso when he heard the news. Only the beverage's golden crema residue was left clinging to the sides of his cup.

Translation: First RP*** the reporter could find says "Yeah, people drink coffee" rather than "how did you calculate the mean weekly caffeine use?"

The poll and the "index" are mediocre at best -- exactly what you'd expect somebody to cook up for a mediocre PR effort. But it's a pretty good return on investment when you can get a once-competent newspaper to repeat "shine a light on the numerous health benefits" on its front page. Hard to see how we're going to fix the science reporting thing when we do such an amateurish job at gatekeeping in the first place.

* Need someone to fax you the lyrics?
** Hi, Margaret!
*** Real Person. It's a Kennedy thing.


Blogger Strayhorn said...

And apparently they missed the Most Important aspect of this story: whether or not you left a tip.

That rated 7 minutes, a moderator, a reporter, and two calls to various participants this morning on NPR.

See? Journalism is not dead. It's just had two Quaaludes with a fifth of cheap gin.

2:24 PM, November 09, 2007  

Post a Comment

Links to this post:

Create a Link

<< Home