Monday, September 18, 2017

Propagating a Bad Statistic

No one truly believes that 1/5 of women will experience a sexual assault in college.  Such a rate would be worse than some of the most unsafe countries on the planet.  If anyone believed that 1-in-5 number, they wouldn't go to college, or send their daughters to college.

Yet here we have The Economist spreading the lie:
ANY sentence containing the phrases “Donald Trump” and “campus sexual assault” could reasonably be expected to conclude with the word “outrage”. Yet when Mr Trump’s secretary of education, Betsy DeVos, announced her intention to “revoke or rescind” directives to universities on handling sexual assault issued by the Obama administration, the move was quietly welcomed by plenty of colleges. The Obama administration’s determination to discourage campus sexual assault—which is suffered by as many as a fifth of women attending college—was well-intentioned, but poorly thought out.
One could argue that technically, The Economist is correct--they did say "as many as a fifth".  They'd be just as correct had they said "as many as 99%", too, so I don't accept the quibble.

The 1-in-5 number is a bad statistic that traveled around the world before the truth even got out of bed.

According to the Washington Post:
In the Winter of 2006, researchers used a Web-based survey to interview undergraduates at two large public universities, one in the Midwest and one in the South. A total of 5,446 undergraduate women, between the ages of 18-25, participated as part of a random sample. The survey was anonymous and took about 15 minutes to complete. (Participants received a $10 Amazon.com certificate for participating.)
My introductory statistics students can pick apart the problems presented in this paragraph; toss in a low response rate, and anyone who uses that information deserves to be mocked.

Politifact tells the same story, with this tidbit tossed in:
"This ‘one in five’ statistic shouldn’t just be taken with a grain of salt but the entire shaker," said James Fox, professor of criminology, law and public policy at Northeastern University.
The author of the Economist article should be ashamed of him/herself.   The Economist should do better.

1 comment:

Ellen K said...

So no validation of the people polled, no verification of their age, sex, status and personal attitudes toward alcohol and drug use. You can make a poll say anything you want it to depending on how you shape it. And yet Big Media goes with the unsubstantiated sexy lead instead of spending a buck on verified polling. There's alot of different definitions of "assault" Some people use it interchangeably with rape. Some consider touching assault. Some consider dirty jokes even when not aimed at them specifically to be assault. Without a legal definition the poll is worthless.