We'd really appreciate it if Al over at Poynter would stop passing bogus story ideas along at his "morning meeting
." Here's the latest:Super Bowl Sick-Out
There does seem to be some truth to one commonly held belief -- lots of people miss work Monday after the big game. MSNBC.com reports:Super Bowl flu is quite contagious and quite common, says one employment productivity expert who has studied absenteeism in the workplace.
Here's a generalization for you: Whenever you see an absenteeism story built around the numbers of a "productivity expert" or "workplace guru," assume it's bogus. Pull it out of the queue and search it until it confesses. Generally, what's happened is that some wise marketing shill has sold a reporter a handful of magic beans and is long gone across the county line enjoying the benefits of a free ad. Let's take the above bit of fiction apart:“We think there’s going to be a widespread impact across the country of Super Bowl-itis — and epidemic proportions in Chicago and Indianapolis,” said Stewart Itkin, vice president of Kronos, a Massachusetts company that solves workforce-related problems. (
Clue #1: "Solves workforce-related problems" how? Mass layoffs? Decimating the board of directors?)
Nonetheless, the mark -- that's you,
bubba -- is set up:Kronos recently surveyed approximately 1,300 adults over the age 18 and asked if they would be coming to work the morning after the game, Feb. 5. Five percent of the respondents admitted that they planned to call in sick. With the U.S. working population numbering about 140 million people, Itkin estimates that "come Monday, there will be 7 million empty cubicles around the country, costing employers hundreds of millions of dollars in lost productivity."
Well, that's pretty simple arithmetic. You can do that at your desk between phone calls. If you assume that self-reported intention translates exactly into action. And that people take a question like "Are you gonna call in sick the day after the Super Bowl?" seriously. And that his guess about productivity has something to back it up. And that ... oh, right. That his survey is a valid measure of something.
Neither Al nor MSNBC says, but Al gives us a link to Kronos' Web site, whence we can find something about the "previous survey" mentioned in the story (according to all of which the problem has more than quadrupled
in two years). Scope the tactics in this press release from 2005:New survey findings suggest that an estimated 1.4 million employed U.S. adults may call in sick to work the day after the Super Bowl. The "Working in America: Absent Workforce" survey of more than 1,300 employed adults was conducted by Harris Interactive® and commissioned by Kronos® Incorporated. Super Bowl-related absence rates may be even higher in parts of New England, the Mid-Atlantic region, and other areas where the contending teams have strong followings.
Where are the numbers for the last assertion? Hey, who cares? It sounds reasonable! Who's going to bother to check out an assertion like this? Which is how we get paragraphs on the news pages like this (yes, Lowell Sun, it's time for you to be embarrassed):One way to better plan for such crises is to automate the tracking of sick time and schedules with software, one type of which is designed by Kronos.
Now, Harris Interactive is a grownup organization with a reputation to protect, so it's not going to lie about its methods. But hang on for the bait-and-switch here:Harris Interactive conducted the online survey on behalf of Kronos Incorporated in the U.S. between December 20 and 23, 2004 among a nationwide cross section of 1,316 full-time employed U.S. adults aged 18 and over. Figures for age, sex, race, education, income, and region were weighted where necessary to bring them into line with their actual proportions in the population. Propensity score weighting
(I don't know that that means, but I'm impressed)was also used to adjust for respondents' propensity to be online. In theory, with probability samples of this size, one could say with 95 percent certainty that the results for the overall sample have a sampling error of plus or minus three percentage points.
(Why be coy? Come right out and say it in practice! At a 95 percent confidence level, a probability sample of that size has a confidence interval, or margin of sampling error, of less than 3 points.) Sampling errors for the results of the following sub-samples: employed adults who have called in sick to work in the past year (792); employed adults who have a disciplinary attendance policy at work (733); and employed males aged 25-44 (300) are higher and vary.
(Sun rises in East.) This online sample was not a probability sample.
See? If you get through all the cautions and hedges in the methods graf, you get to the line that knocks it all into a cocked hat. This isn't a probability sample
, and that means you can't draw any conclusions about the wider population from it. Kronos has the free ad, you've gotten to make some stupid generalizations:To wash it all down, we'll guzzle some $12 million worth of beer alone.*
and nobody's the worse for wear. Unless you're really tired of seeing made-up stuff crowding news out of the newspaper.
* Copydesk math quiz: Schnucks is selling Miller Lite for $15 a case. Based on that price, how many beers do "we" get to guzzle on the day of the big game?