I was just jerkin your chain about the research stuff. I know you already do a huge amount of it.

Sure, a President's approval ratings go up during a crisis, as long as they're perceived as handling it, and that approval is usually honestly earned. But terrorism presents unique opportunities for manufacturing threats to manipulate public approval. Terrorists are the monsters who hide in plain sight, perhaps living next door to you, or maybe that swarthy man standing in line at the airport, or even...hiding under the bed! And all it takes to make millions tremble is to order the terror alert level raised. You even get a pass on refusing to provide any details on the grounds of "national security".

In other words, terrorism presented Geppetto Rove with a political tool so powerful and yet so cost-effective that it's beyond belief to think he wouldn't have used it. And by the standards of American politics I'm not particularly slandering Rove--any political adviser in his position would have been tempted past human endurance to grab the opportunity.

You're right that only two of the bumps in the graph are large, naturally because they're associated with major events that boosted Bush's popularity independently of the terror alert levels. This is probably a different set of data from the Cornell study, but applying their 3% result, you can see the correlation at a finer scale: Raising the alert level provides a boost of a few percentage points, lowering it drops it by a similar amount.

Often, anyway--there are other bumps on the curve not associated with terror alert levels. Beats me. You'd have to correlate with all sorts of events in the outside world to be really confident that you've isolated an actual causal relationship with terror levels. Take it with a grain of salt.