Hey, do you remember that oft cited Newsweek article from 1986 that proclaimed that the chances of a 40-year-old single, white, college educated woman getting married were less than her chances of getting killed in an act of terrorism? It turns out it was wrong. From a recent retraction of that article:
Twenty years later, the situation looks far brighter. Those odds-she'll-marry statistics turned out to be too pessimistic: today it appears that about 90 percent of baby-boomer men and women either have married or will marry, a ratio that's well in line with historical averages. And the days when half of all women would marry by 20, as they did in 1960, only look more anachronistic. At least 14 percent of women born between 1955 and 1964 married after the age of 30. Today the median age for a first marriage--25 for women, 27 for men--is higher than ever before.
It would have been nice in this retraction to see some discussion of how the demographic research discussed in the retracted article managed to make such wildly "pessimistic" predictions. (Were there bad assumptions built into the predictions? Was it a weird lump of data for the purposes of predicting marriage trends? Are there any lessons here to help us tell whether future such predictions are worth trusting or worth binning?)
But, I suppose we must remember that we're talking about Newsweek here. Perhaps the strongest selling point, from the point of view of turning demographic research into a cover story, is that the statistics "felt true", at least to someone making judgments about what to print. For the record, feeling true is not sufficient to meet a scientific burden of proof. Subjective experiences don't make a claim correct -- even if people's interaction with their own experience (whether getting anxious about low probability of marrying or deciding not to count on marriage as a certainty) changes in light of a statistic they are told is correct. Probably there's a lesson here about how people think about probabilities in their everyday lives, but as far as I can tell the bigger lesson is not to count on the scientific goodness of reported "trends" that play into pre-existing societal anxieties.
(Hat tip: Pandagon1>
I stashed that issue of Newsweek away in my office for future use whenever I teach statistics. It a perfect bad example. The results are patently ridiculous, but lots of people took it seriously and were traumatized. I wrote a post about how a good statistics course is an important part of any person's self-defense repertory. Don't leave school without it!
Statistics: Math for self-defense includes several links to posts about the Newsweek debacle at Feministe, the Wall Street Journal, and Newsweek itself.
The longer article in the actual magazine provides an explanation for the faulty conclusion. I don't recall the details offhand, but I think it had to do with extrapolating from fairly old statistics., i.e., in the 50's, if you weren't married by the time you were 30, there was only about a 1% chance you'd ever get married. This ignores the fact that in the 50's, 99% of the people who did get married got married before they were 30.