It's common to dismiss numbers that prove somebody else's point as "lies, damned lies and statistics."
That kind of flip refusal to engage with data is a sure sign of someone who doesn't want to think too hard. Or someone who is deeply suspicious of anything that looks like math.
Numbers are numbers. If they're properly gathered and accurately reported — that's key of course — they just don't spin, obscure, imply, connote, the way that the English language can. Obviously people are tempted to emphasize numbers they like and downplay numbers they don't like, say, in press releases and things like that; but if you take the time to go back and look at the source data, and get to decide for yourself which numbers are important, you can get at something approaching truth.
So I reject the glib "statistics = lies" equation.
But then something comes along that makes me want to title a post "Lies, damned lies and statistics."
James Taranto reports:
Amid all the gloomy economic news, the New York Times brings us an encouraging report on social trends:
"The number of black children being raised by two parents appears to be edging higher than at any time in a generation, at nearly 40 percent, according to newly released census data. . . .
"According to the bureau's estimates, the number of black children living with two parents was 59 percent in 1970, falling to 42 percent in 1980, 38 percent in 1990 and 35 percent in 2004. In 2007, the latest year for which data is available, it was 40 percent."
What accounts for the turnaround? The Times explains:
"Demographers said such a trend might be partly attributable to the growing proportion of immigrants in the nation's black population. It may have been driven, too, by the values of an emerging black middle class, a trend that could be jeopardized by the current economic meltdown.
The Census Bureau attributed an indeterminate amount of the increase to revised definitions adopted in 2007, which identify as parents any man and woman living together, whether or not they are married or the child's biological parents."
The problem of illegitimacy and broken families had seemed intractable for decades, but the Census Bureau has been able to make a significant dent in it, at virtually no cost to the taxpayer, merely by redefining the word parents.
I have to remind myself that it's not the numbers that are lying. What's wrong here are definitions. Which has to do with language, with words.
Although it's a handy reminder that language and mathematics are perhaps inextricably combined. For isn't classifying items into sets a basic operation of mathematics? And how do you do that without naming them?
In any case, shame on the NYT for reporting the "good news" uncritically in the lede, and burying deeper in the story, with almost comical understatement, the sleight-of-hand, the redefinition of terms — a really heinous one, a damned lie if I ever saw one.
(One hopes that the Census Bureau was still asking the right questions, so that the numbers are in fact there, somewhere, buried in the data, and that it was only in turning the numbers into the words of a press release, of a report, that the fudge occurred. Otherwise — if all they asked was, "is there a man and a woman residing at the same address as this child?" — we'll never know, will we?)
bearing blog 