How Statistics Get Distorted — An Example

On a legal academic discussion list, someone said — quoting the Brady Campaign that “In two thirds of battered women’s households that contained a firearm, the intimate partner used the gun against the woman, usually threatening to shoot/kill her (71.4 percent) or to shoot at her (5.1 percent).” Given that there are a lot of women who are battered by intimates — the National Crime Victimization Survey reported 900,000 in 1998 — that sounds like a vast number of death threats.

But if you actually look at the cited study (which is limited to California), you’ll see that it reports on women staying in shelters for battered women. Women staying in battered women shelters aren’t just the typical battered women — they are women who feel so endangered that they decide to flee their own homes. It stands to reason that many of them would have been deliberately threatened with death or serious injury at some point, which may be what led them to flee in the first instance; they are probably more likely than the typical person, including the typical domestic violence victim, to have been threatened in an especially serious way. And it stands to reason that in those homes where a gun is present, and the woman’s partner is willing to deliberately threaten the woman enough to cause her to flee, at least one of the threats will have involved the gun.

So the study (even if it’s sound on its own terms, a matter on which I won’t opine here) doesn’t actually speak about battered women’s households — it just speaks about the households of those California battered women who felt so endangered that they go to a battered women’s shelter.