Tuesday, January 19, 2010


Sam Smith - Thanks to the work I do, I'm constantly looking at percentages and other such figures developed after much, and often expensive, research. And as I do so, the ghost of Alice Darnell repeatedly arises, reminding me - as she did teaching my 12th grade math class - that a result can be no more accurate than the least accurate figures used to derive it.

Yet polling firms, government agencies, and others trying to prove a point to justify their existence, repeatedly tell us things like, "Candidate A received 46.7% in our latest survey." Further down, however, you learn that the poll is accurate only within 3-4 percentage points.

In the notoriously inadequate world of school testing, it can get even worse. The DC school system tells us, for example that 47.62% of fourth graders are "proficient or advanced in reading," a number that doesn't reflect how many students were sick on the day of the test or to what degree questions with multiple choice answers reflect actual reading skill or how many students read better than they take tests. But the score certainly reflects a test company's ability to con its clients and the lack of math skills of those running the DC school system.

My own research has found that 98.12% of the percentages I come across in news stories don't deserve their extra decimal points and journalists should round them up or down in a well-deserved attack on digital deficit disorder.