News

Loading...

Tuesday, October 19, 2010

How to (Subtly) Lie with Statistics

Look, I suspect the main thesis of this article is spot-on: Americans are less connected to their communities than they were a generation or two ago. But note the bogus maneuver executed in this paragraph:

"Moreover, the current that Putnam observed has, according to more recent studies, only intensified in the last decade. One study found that Americans had one-third fewer nonfamily confidants than they had 20 years earlier, and 25% had no one in whom to confide whatsoever. Another study of 3,000 Americans found that on average they had only four close social contacts, but these included family members like one's own spouse."

The theme of the paragraph is how Americans' isolation has "intensified in the last decade." And then several studies are cited to back that point. But, in fact, the last study cited, on its own, is totally irrelevant to the point at hand. It only describes the current state of American society, and gives us absolutely no traction for answering the question "Does this state of society represent an intensification of the trend Putnam observed?" Perhaps a decade ago, Americans had, on the average, only two close social contacts, so that this statistic represents a reversal of the "current" that Putnam observed. Now, I'm not saying that is the case: I'm only pointing out that a statistic that is essentially a static snapshot cannot tell us anything about the direction of a dynamic process, and that this author either was too dull to realize this or was deliberately bullshitting us by including it.

1 comment:

  1. Gene,

    Supposed that the Galactic Cyclopedia reports that the human population in 2020 C.E. was zero. Doesn't this tell us something about the diachronic process of human evolution?

    ReplyDelete