Are terrorists statistically significant?

tsa_securityHere’s another discussion relevant to the earlier post that DO SOMETHING is not a helpful government response to the current terrorismscare:

[T]he key point about identifying al-Qaeda operatives is that there are extremely few al-Qaeda operatives so (by Bayes’ theorem) any method you employ of identifying al-Qaeda operatives is going to mostly reveal false positives.

(From Matthew Yglesias via Tyler Cowen ONCE AGAIN, I think I’m now Tyler’s full-time RA).

How does this relate this to our usual statistical analysis? Proving someone is a terrorist is analogous to proving a nonzero effect that confirms an economic theory. We allow a rate of false positives of 5 percent (“statistical significance at the 5 percent level”) for showing that, say, good institutions have a positive effect on development. The false positives do not automatically swamp the true positives, because the true effects of  one thing on something else are not as rare as terrorists.

To have a low rate of false positives, we have to accept a high rate of false negatives. But we don’t care about false negatives. You failed to show an effect of your favorite magic ingredient X on development? Too bad, the burden of proof is on YOU if you want to add your ridiculous theory to the existing development knowledge.

Contrast airport security, where we DO care about false negatives (i.e. failing to detect a terrorist). To reduce false negatives even more (as everybody is demanding ), we would have to accept MORE false positives. This would swamp even more the rare genuine terrorists.

Yglesias used a hypothetical rate of false positives of 0.1 percent in his discussion of screening 15 million British Moslems. Of course, TSA makes it much worse by screening each and every of the 800 million airline passengers annually in the US -- including my 80 year old mother whose only suspicious behavior is hiding her handbag in fear of NYC purse snatchers. A false positive rate of 0.1 percent times 800 million means that false positives would be 800,000 people.

Have you seen 800,000 terrorist suspects milling around at airport security? No, I haven’t either. So the true TSA false positive rate must be even lower, which must mean the false negative rate must be a lot higher than the TSA would like to admit (as confirmed by audits). Intensified universal screening cannot possibly work: QED. (For useful alternatives, consult the people in the know.)

As Shakespeare once said about TSA:

It is a tale told by an idiot, full of sound and fury signifying nothing.

OBLIGATORY AID PARALLEL TO TODAY’S APPARENTLY UNRELATED NEWSWORTHY TOPIC: the Do Something approach in aid has not been a great success either. Although it is still popular in aid and social activism, as illustrated by the nearly 300,000 followers on Twitter of @DoSomething, who wrote the following “Tweet”:

Its really easy to be a critic. Its really hard to be a do-er who actually makes stuff happen.

Stuff happen like click on a non-binding poll on their web site whether unnamed state legislators who don’t check web sites should pass laws against texting while driving. That may be easier than being a critic.

UPDATE: announcement today that TSA will piss off 14 mostly Muslim countries by subjecting fliers from those countries to the US to universal invasive screening. Thank goodness the terrorists are so dumb they would never think of flying from ANOTHER country besides these 14!