Poll Confusion Made Less Confusing

Reader Cristine Barbour, a political scientist at Indiana University, notices that the USA Today/Gallup poll sampled its wares from Democrats and Democratic-leaning independents, NOT likely voters, which is the universe from which the CBS News / New York Times poll was derived.

In fact, the CBS/NYT poll also has Clinton up (well, tied -- 45 to 44) among its sample of Democratic identifiers (registered voters who vote Democratic with leaners pushed), with Obama leading among likely voters. So the divergence is artificial (although still interesting).

Obviously, the much larger pool of Democratic identifiers is more of a weathervane sample, apt to support Clinton when the noise is unfavorable for Obama and vice versa. The smaller likely voter samples suggest that Obama is much more durable among those who have or will bother to show up at the polls.