In my February 10 post, I noted that Gallup’s 57 percent approval rating for Bush in their February 4-6 poll was way out-of-line with other recent polls and was likely driven by the lopsided Republican party ID advantage (+9) in their sample, rather than any real world shift in public sentiment.
Now Gallup has a new poll, conducted February 7-10, and lo and behold, Bush’s 57 percent approval rating has nosedived to 49 percent, in the same neighborhood as other recent polls. It’s just a wild guess, but I shouldn’t be surprised if this lower rating for Bush was also accompanied by the disappearance of that big Republican party ID advantage (paging Steve Soto!).
You’d think that their first poll’s heavily Republican sample and outlier status relative to other polls would make the Gallup folks cautious in interpreting Bush’s approval rating drop in their second poll as representing a real political trend. Think again! Gallup, in fact, contructs a whole political story to fit their data, starting with their title “From Public’s Perspective, Past Week Not Good One for Bush: Approval ratings drop” and continuing thusly:

The new poll, conducted Feb. 7-10, shows his approval rating is back down to 49% as the news focus has shifted to his proposed federal budget and his plans for changing the Social Security system.

Could be. But an alternative, more parsimonious hypothesis is this: Bush never had a popularity spike to begin with; little changed over the period covered by Gallup’s two polls and Gallup’s swing against Bush in their latest poll is an artifact of their changing sample composition, not the product of a real shift in public opinion.
But that wouldn’t fit with the Gallup philosophy: reality’s job is to explain Gallup’s data, not the other way around.