John Allen Paulos in The New York Times:
In the realm of public policy, we live in an age of numbers. To hold teachers accountable, we examine their students’ test scores. To improve medical care, we quantify the effectiveness of different treatments. There is much to be said for such efforts, which are often backed by cutting-edge reformers. But do wehold an outsize belief in our ability to gauge complex phenomena, measure outcomes and come up with compelling numerical evidence? A well-known quotation usually attributed to Einstein is “Not everything that can be counted counts, and not everything that counts can be counted.” I’d amend it to a less eloquent, more prosaic statement: Unless we know how things are counted, we don’t know if it’s wise to count on the numbers.
The problem isn’t with statistical tests themselves but with what we do before and after we run them. First, we count if we can, but counting depends a great deal on previous assumptions about categorization. Consider, for example, the number of homeless people in Philadelphia, or the number of battered women in Atlanta, or the number of suicides in Denver. Is someone homeless if he’s unemployed and living with his brother’s family temporarily? Do we require that a women self-identify as battered to count her as such? If a person starts drinking day in and day out after a cancer diagnosis and dies from acute cirrhosis, did he kill himself? The answers to such questions significantly affect the count.