A graph of an alleged correlation between Bush's approval ratings and the timing of terror alerts, commonly attributed to a blog called JuliusBlog, has been circulating around the internet and progressive discussion boards such as Democratic Underground since last summer. Now, in response to new comments from former Homeland Security Chief Tom Ridge, the mega-blog Eschaton has linked to this graph (though via a guest blogger and not by Atrios), so I suspect it will get a renewed rush of attention.
Those who promote the graph claim that it demonstrates proof that the Bush administration politicized terror alerts. Supposedly when Bush's approval ratings drop, the graph shows that he issues a terror alert.
But the graph actually shows no such correlation.
The graph does indicate that Bush's approval ratings were almost constantly in declining mode. So mere chance dictates that most terror alerts were issued during periods of declining approval ratings.
Only during three brief periods shown (perhaps four), did Bush's approval ratings rise: after September 11, 2001; at the onset of the Iraq War; at the capture of Saddam Hussein; and possibly after the first anniversary of the September 11 attacks. Despite the brief time period of these rising approval ratings, two of the six "orange alerts" on the graph came during these brief periods of rising Bush approval ratings. The hypothesis that Bush used terror alerts to boost sagging approval ratings cannot explain this.
One possible explanation for the common misinterpretation of this graph is that many are mistaking the wiggles of the red line for actual Bush popularity changes, though they are almost certainly statistically-insignificant noise in the data. Even so, there doesn't appear to be a correlation, so perhaps the true explanation is that people are simply seeing what they want to see.
But take away the data about approval ratings and simply examine the timeline of terror alerts. This shows that , come summer of 2004, in the run-up to the Presidential election, the frequency of terror alerts dramatically increased. In my view, this is an illustration (though of course it does not prove) that Bush indeed used terror alerts politically. Not to boost sagging approval ratings, but to maintain a climate of fear in this country, since fearful people rally around their leaders, no matter who they happen to be.
This became more necessary in 2004 than in 2002 and 2003, not just because of the impending election, but because memories of the September 11 attacks were fading. In fact, the gradual fading from memory of the horror of those events may well be an explanation for the slow, steady decline in Bush's approval outside of the three or four dramatic events mentioned above.
The new comments from Tom Ridge, who claims that his reluctance to raise terror alert levels was often overruled by administration officials, reinforce the idea that the Bushies politicized the terror alert system prior to the 2004 election. But the graph of Bush's approval ratings versus the timeline of terror alerts does not.
UPDATE: Via Carpetbagger Report and Kevin Drum, I see that a quantitative analysis of a possible correlation between Bush's popularity and terror alerts was actually performed by a Cornell University sociologist last fall. Through a complicated statistical analysis, the sociologist finds a small tendency for Bush's popularity to temporarily rise after government-issued terror warnings. The magnitude of the effect he found was 2.75% on average. This is a different effect than the one frequently cited by promoters of this graph, and quite small as well, even if the sociologist has managed to accurately control for all the other effects surrouding Presidential popularity. But it is a reminder of how thorough statistical analysis can be more useful in demonstrating a correlation than vague graphs.
UPDATE #2 (6:39pm): I added a paragraph towards the end of the article to clarify a point: the one beginning "This became more necessary... ."