Except It Doesn't

Richard Morin, "What's Black and White and Red All Over?" Washington Post, June 15

More ink equals more blood, claim two economists who say that newspaper coverage of terrorist incidents leads directly to more attacks.

It's a macabre example of win-win in what economists call a "common-interest game," say Bruno S. Frey of the University of Zurich and Dominic Rohner of Cambridge University.

"Both the media and terrorists benefit from terrorist incidents," their study contends. Terrorists get free publicity for themselves and their cause. The media, meanwhile, make money "as reports of terror attacks increase newspaper sales and the number of television viewers."

The researchers counted direct references to terrorism between 1998 and 2005 in the New York Times and Neue Zuercher Zeitung, a respected Swiss newspaper. They also collected data on terrorist attacks around the world during that period. Using a statistical procedure called the Granger Causality Test, they attempted to determine whether more coverage directly led to more attacks.

The results, they said, were unequivocal: Coverage caused more attacks, and attacks caused more coverage -- a mutually beneficial spiral of death that they say has increased because of a heightened interest in terrorism since Sept. 11, 2001.

No particular surprise that this was picked up by a number of blogs (Cliff May at The Corner, Insty, Outside the Beltway, Dr. Sanity, and CBS's Eye, among others) whose predisposition to believe trumped a stated (or, in the case of Doc Sanity, a demonstrated) lack of understanding of statistical methods. But that doesn't really excuse a complete unfamiliarity with studies using such methods or of the general run of empiricism itself; the required response to any such study is skepticism tinged with curiosity. (I'd also argue that, pace everyone above, the standard intuition of the issue should that it's wrong, not that it's self-evident.)

Let's begin with our source: like too many journalistic forays into "science" we take an interesting or provocative study and clothe it in the panoply of modern science with a choice phrase or two; here it's "the Granger Causality Test". Our first objection is that Morin had an obligation to explain what the Granger Causality Test is, not what someone planned to do with it, unless what they planned to do with it offered some proof in the success or failure of application. If I'm testing whether dropping a baseball onto my sister's head from the top of the swingset will make her yell for Mom (like that first kiss, you never lose the warm feeling you got from your first scientific foray), reporting my findings is sufficient. If my purpose is to determine whether news reports "Granger cause" terrorist attacks, then regardless of the results I need to know what the hell "Granger-cause" means. Absent that--and we are--it is not sufficient for the reader to say, "I don't know anything about statistics, but the guy who came up with that one won a Nobel, so this must be right", or to use Doc Sanity's term "objective proof".

We don't have to search very far, nor prise meaning from any mathematical expressions before we sense there's a bit of a problem. A quick Google search will introduce you (as it would have introduced Mr. Morin) to Bent E. Sørensen, Ph.D., lay professor of economics at the University of Houston. And Dr. Sørensen will be kind enough to explain:

Often you will have that xt Granger causes yt and yt Granger causes xt. In this case we talk about a feedback system . Most economists will interpret a feedback system as simply showing that the variables are related (or rather they do not interpret the feedback system).

Sometimes econometrians use the shorter terms “causes” as shorthand for “Granger causes”. You should notice, however, that Granger causality is not causality in a deep sense of the word. It just talk about linear prediction, and it only has “teeth” if one thing happens before another. (In other words if we only find Granger causality in one direction)....
Granger causality measures whether one thing happens before another thing and helps predict it - and nothing else. Of course we all secretly hope that it
partly catches some “real” causality in the process. [emphasis mine]

So-oooo, in other words, the "unequivocal" results mean that Frey and Rohner are not the first men to achieve metaphysical certainty; they found that news stories of terrorist acts are related to the number of terrorist acts. For my part, I'm willing to agree that this much seems intuitive, in the "well, duh" sense. In fact Frey and Rohner's abstract states that terrorist acts and media coverage "mutually Granger cause each other", giving the firm support of two legs of a three legged stool.

But "deep causality?" Whoever's pitching that one it's No Sale.

The second area where natural skepticism should be applied is the methodology of the study. Again, it should be fairly automatic to raise questions when both the "Media" and the "Terrorists" are aggregates, especially when the media part of the equation consists strictly of Western media. (Isn't local coverage more crucial to most terrorist attacks?) There's also the little matter of defining "terrorism", which has been left to those same Western papers.

And there are other curiosities. "Fame and power" are listed as two terrorist motivators, while "sensationalism" is a reward for the media. There's also the mention of an "interesting article by Nelson and Scott (1992)" which covered the same question for 1968-1984 and concluded that media coverage did not increase terrorism, to which the current authors add that their study tests "whether this conclusion still holds in today's more globalised and media-covered world". But then, this is also a world where the government of the US has invaded two countries using terrorism as the grounds, and where four major attacks on Western cities have occurred in the past five years. My own opinion, sans formula, is that the world was "globalized and media-covered" sufficiently in 1968 that those attacks would have been big news back then, too.

Of course the idée fixe on the other side of the looking glass is that the MSMSMSMSM actively encourages the Moslem hordes, not for the sensationalist buck but out of its unstinting anti-Americanism. That's something a bit more than Frey and Rohner, with their suggestion of reducing the sensationalism of news coverage, conclude, but then nobody seems to have paid close attention to anything they had to say anyway. If the US media ever stopped sensationalizing terrorist attacks you know who'd be calling them on that, don't you? Or should we be running a Granger causality test on the correspondence between right-wing blowhards and disastrous US foreign policy?