Jonathan Bernstein picks up on one of my hobbyhorses:
Here's the thing. Barack Obama isn't as popular now as he was in January 2009. This is not exactly a little-known fact; indeed, we fortunately have some really good indicators of exactly how popular Obama is overall, and they're not all that obscure.
What this means is that sloppy journalists can get endless mileage from picking out any subgroup in the nation and finding out that, gee, Obama has lost popularity there!....To know whether any of these stories is actually news, it's absolutely necessary to compare Obama's decline within the group in question to his overall decline. If it's more, then you have something; if it's the same or less, then you're at best illustrating how an overall decline works within that subgroup.
Roger that. This usually bugs me most during the post-election recap season. In 2008 the media was full of breathless reports about how Obama gained support among married women or McCain lost support among Hispanics or some such. But of course they did. In 2008 Obama did a lot better among all voters than Kerry did in 2004, and McCain did worse than Bush. So it stands to reason that Obama also did better among most demographic groups and McCain did worse. In 2008, for example, several writers suggested that Obama did especially well among churchgoers, but in fact he didn't: he performed about 9 points better than Kerry overall and about 10 points better among churchgoers. There was nothing to it.
Anyway, this is just another example of "compared to what?" That's a question that should be on everyone's minds a lot more than it usually is.