Wednesday, 23 July 2003

Lott lie?

Wyeth alerted me to this post in which he says:

John Lott—whose survey evidence for More Guns, Less Crime disappeared in a mysterious computer hard drive crash*—is trying to make the case that an armed Iraq is a safe Iraq:
“Yet, despite Iraqis owning machine guns and the country still not under control, Defense Secretary Donald Rumsfeld pointed out that Baghdad is experiencing fewer murders than Washington, D.C., where handguns are banned.”

Let’s forget for a moment whether it is good politics to tell the American people that you want Iraqis to have as many guns as possible at a time when our soldiers are being killed every day by those guns.

Let’s focus on a smaller point—are John Lott’s statistics even accurate? Is the murder rate in Washington DC higher than the murder rate in Baghdad?

Now, it’s possible to know anecdotally what the approximate murder rate is without having detailed statistics available from a central agency. Presumably someone in Baghdad is still making out death certificates, and deaths are being investigated. So, if there are fewer than 262/365 (0.72) murders per day on average (i.e., a murder is only reported every other day, or less often), the murder rate is lower in Baghdad than in Washington.

You can reasonably argue about the causal mechanism; I suspect murder rates could be lower for more complex reasons than “everyone’s armed” (for example, many of the sociopaths who would otherwise be inclined to commit murder were likely Saddam Fedayeen recruits and have been wiped out by the 3ID and others, or maybe it’s just part of the post-war adjustment to a new government by the population). But I’m not sure quoting a statement by a senior administration official is, in and of itself, a lie; at worst, it’s disingenuous support for one’s own position, particularly in the absence of compelling evidence to the contrary.

For example, if I say “Bill Clinton says he did not have sex with that woman, Miss Lewinsky,” that doesn’t make me the liar; it does make Bill a liar, unless you want to quibble over the definition of “sex.” In February 1998, it would have been reasonable for me to take Clinton’s statement at face value. Today, even in the light of compelling evidence to the contrary, unless I say “this proves Bill didn’t have sex with Monica” I’m still not a liar.

So, unless someone has statistics showing that the current murder rate in Baghdad is greater than 0.72 people per day (which translates to just over five murders per week), John Lott isn’t necessarily a liar. It is, however, distinctly possible that Lott is wrong. Now, if Lott is subsequently informed that Rumsfeld is factually incorrect, yet continues to repeat the claim, then it would be reasonable to claim he is lying.

Again, a review for those of you just joining us here at SN: lying requires foreknowledge that you are making a factually incorrect statement. Being wrong just requires that the statement being made (or quoted) is factually incorrect. In other words, lying requires intentional deception on behalf of the speaker in addition to factual incorrectness.

James Joyner (in trackback below) makes an important point:

Of course, univariate analysis is silly. Baghdad and Washington are hardly comparable cities. Indeed, one would expect a lower homicide rate in a police state than in a free society.

Indeed. And, that would be a worthwhile critique of Lott’s analysis, which gets to the whole “causal mechanism” thing I discussed above. The best I can say for Lott (if you accept his claims about the dispensation of the survey data, which I find dubious but not entirely improbable) is that he’s a sloppy social scientist—albeit perhaps not an not extraordinarily sloppy one, given the pure sludge that often is passed off as strong evidence in many peer-reviewed journals.

* Lott’s missing data only affects a small part of his overall argument; it may speak to his overall credibility, but the vast majority of his data is available and has been analyzed by other scholars.