Throwing the terror switch

Well, sure. (The terrorism rays are a particularly nice touch, don't you think?) As the story tells it, "terrorists" could indeed hire or develop the hackery to let them hijack ships. So, depending in part on when your moviegoing tastes were formed, could pirates, Soviet Russian communists, and 007-worthy mega-villains. But that's not how you get to the top of the Fair 'n' Balanced homepage. For that, you need you some terrorists.

Which is one of the important takeaway points of this very nicely executed experiment-within-a-survey* from the Pew Research Center. Another poll on what Americans think about gubmint surveillance** would be a fairly small contribution, but one that manipulates the question wording to see if throwing the "terror" switch makes a difference -- now we're having fun.

If you weren't familiar with the idea of error introduced by question design, listen and attend. Pew asked a simple "would you favor or oppose" question about "government surveillance programs," but it was asked in 16 conditions: four levels of "what" (collecting metadata about phone calls or emails, collecting recordings/text of phone calls or emails), two levels of "how" ("with court approval" or nothing) and two levels of "why" ("as part of anti-terrorism efforts" or nothing). So a sample of 2,002, which would be painting the lily for most yes-or-no polls, populates all 16 cells quite effectively.***

The results are striking but hardly surprising. Respondents don't distinguish phone calls from emails in any meaningful way, but everything else makes a difference. My only complaint, indeed, is that Pew doesn't do any basic significance testing, but that's the sort of thing you can do on your own -- so let's.

Take the first table, "Mentioning court approval, terrorism increases NSA support." Each of those conditions includes half the sample, so just multiply the percentages by 10 to hairball the number of cases. US adults are more likely to approve of surveillance if it's "with court approval," and that difference is statistically significant: χ2 (df = 2) 33.73, p < .001, v = .13.

Is that a bigger or a smaller difference than throwing the terrorism switch? A little, but not much. Mentioning "anti-terrorism efforts" produces an equally significant but very slightly smaller effect, χ2 (df = 2) 22.01, p < .001, v = .105. Phone vs. email is a wash, but metadata vs. recordings is significant, χ2 (df = 2) 7.79, p = .02, v = .06.

Without the spreadsheet itself, you can't get deeply into the ANOVA weeds, but a few other comparisons do suggest themselves. As long as the "terror" switch is thrown, respondents don't care whether or not a court has approved gathering metadata (p = .489, or about one chance in two that the apparent difference came about by accident). But they're very concerned about actual recordings or text captures: the court's approval makes a significant (p = .005) difference in approval there.

Numbers are fun, but they're fun because they provide a degree of confidence in things that look like commonsense conclusions. It makes sense that the "take my habeas corpus -- please" response is stronger when somebody whispers "terrorism," but if you want to complain about it, it's nice to have a well-designed experiment to support the case.

It's no wonder that the outrage story of the day at the Fair 'n' Balanced Network is about whether the craven librul media have bought the Kenyan Muslim usurper's audacious suggestion that this year's bogus scandals are "phony." Benghazigate is a phony scandal because it has nothing to do with Mideast policy, or the deaths of Americans serving their country, or anything substantive; it's about whether "terrorism" was said loudly, frequently and fearfully enough. (For the record, Fox itself was eagerly blaming the famous YouTube video days after the administration had said "act of terror," but language philosophy is for a different post.)

How and when you call "terror" on a case is a matter of indirect securitization. You can't see the referent object directly, but once you say "terror," you've identified it, because everybody knows the only reason we have terrorists is that they hate us because they hate freedom. Unlike your tanker-hijacking international mega-criminal, who can be brought down by your Bruce Willis or your Sean Connery, terrorism is an existential threat beyond the bounds of the established system, so you might as well hand over your emails for the duration. We'll tell you when you can have them back.

It's always nice to find some hard evidence to support theory in real life, so yeah: When you can call terrorism on some event, you substantially increase your chances of calling security on it, and when you do that, you give yourself a role in saying when we can all come out of the fallout shelter and have our habeas corpus back. It really does matter how, and how frequently, and how enthusiastically, journalists throw the terror switch.

* Singled out by regular reader Garrett, to whom thanks.
** That'll be "snooping," of course, at certain networks 
*** For whether a nonrandom sample would have been equally appropriate, see Basil et al (2002)