Wednesday, December 12, 2007

Journalists are so awesome with statistics, part two

From The New York Times:

When Mr. Nutter takes office on Jan. 7, he will face a crime wave that has left at least 355 people dead so far this year and that gave Philadelphia the highest homicide rate of any big city in the country last year, with 406 killings — more per capita than even New York City, which has six times the population.

The whole point of a "per capita" (or "per person") statistic is that it doesn't matter what the population is. If a town of 10 people has one homicide and a town of 100 has 10, each has .1 homicides per capita. It's the same principle as the homicide rate, except that's usually calculated per 100,000 people -- consequently, since the writer already said Philadelphia has the highest homicide rate, it's redundant to say it has more murders per capita than New York does.

At first I thought the reporter meant to say that Philadelphia has a higher absolute number of homicides than the Big Apple, even though it has one-sixth the population. That would mean its per-capita homicide count is more than six times New York's. But that's not true -- a few weeks ago, NY already had 433. It's amazing that two big cities could be so close in count when they're so far apart in population, so the NYT's point stands, but it's obvious a math major didn't write that sentence.

Finally, Giuliani's crime cleanup left New York incredibly safe for a big city, so some context would be nice. If I were to rewrite the sentence, I'd either leave out New York altogether, or put a period after "406 killings" and add: "New York City, often considered a model of successful crime control, has a homicide rate about one-sixth of Philadelphia's."

0 comments: