The poll numbers these days are a big tease. They are promising an enormous Democratic blowout while at the same time showing pretty much a dead heat.

The latest entry is a Pew Research Center poll showing that fewer people call themselves Republicans these days—fewer than at any time in 16 years of polling by the Center. The Pew figures, released March 20th and based on interviews with 5,566 registered voters in January and February, show 27 percent identifying themselves as Republican, down 6 percentage points since 2004.

Overall, when independents leaning to one party or the other are counted in, Pew finds 51 percent Democrats, 37 percent Republicans and 12 percent independents not leaning either way.

Poll numbers aside, there also has been enormous enthusiasm on the Democratic side, with massive turnouts in primaries and caucuses in state after state. On its face, that suggests Democrats are likely to crush Republicans on all levels. So how come the general election trial heats show Barrack Obama and Hillary Clinton running about even with John McCain?

The Real Clear Politics average from March 5 to 19th shows McCain ahead of Obama by 0.8 percent and ahead of Clinton by 0.7 percent. In four polls since March 14, they show Obama ahead of McCain by 2 percentage points on average and Clinton ahead of McCain by 3 points. No runaways here. The question is, why not?

One factor may be that McCain has started to look presidential after clinching the GOP nomination March 3rd. A second may be that the tough fight between Obama and Clinton is taking some of the glow off both candidates. Both these make sense, to an extent.

But there’s a third possibility also, which is that the samples in some polls are too heavily Republican for one reason or another. That happens from time to time routinely. This year, however, it may be that some pollsters, working from apparently dated perceptions, are assuming there should be about as many Republicans in their sample as Democrats. If so, that would be a pretty gross error, according to the new Pew figures. For reporters and editors, there’s an obvious way to deal with this question: They should ask the pollsters how their samples divide in terms of Republicans, Democrats and independents. And the pollsters should give out that information without having to be asked.

As far as the Pew survey goes, I’ve focused here on the decrease it finds in GOP affiliation. There’s a lot more in this survey, including how changes in affiliation play out in states that were closely contested in 2004, and in red and blue states in general.