Daily Kos Elections Morning Digest: One chart explains partisan self-ID vs. party registration

Want the scoop on hot races around the country? Get the digest emailed to you each weekday morning. Sign up here. Leading Off:

' MA-Sen: Recently, I wondered about a discrepancy in the partisan composition of two MA-Sen Democratic primary polls: one from PPP, the other from MassINC. While the wording in MassINC's toplines PDF was a little confusing (they referred to all of their demographic categories as "questions"), it turns out that they were using party registration figures while PPP, as always, simply relied on partisan self-identification from its respondents. As you'd expect, these two metrics can yield very different results, and in a separate item, MassINC shows just how divergent they can be:

		OVERALL	PARTY REGISTRATION
		Overall	Democrat	Republican	Unenrolled / Other
Generally speaking, do you usually consider yourself a Democrat, a Republican, an Independent, or something else?	Democrat	36%	80%	1%	13%
	Republican	15%	4%	76%	8%
	Independent	46%	15%	16%	75%
	Something Else	2%	1%	6%	3%
	Don't Know / Refused	1%	1%	0%	1%

Using data from an October poll, MassINC compared partisan self-ID (the column highlighted in gray) with actual registration data. As you can see when reading horizontally, only 80 percent of self-identified Democrats (who comprise 36 percent of respondents) are actually registered as such, while 13 percent are independents. Somewhat similarly, 76 percent of Republicans are officially members of the GOP while 8 percent are indies and, weirdly, 4 percent are enrolled Democrats! (Only 15 percent of the sample were self-ID'ed Republicans, though, so we're talking very tiny numbers.) Meanwhile, about three quarters of self-professed independents (who make up close to half of the poll) are members of third parties or are "unenrolled" in either party, but a full quarter are actually registered Dems or Republicans.

It's for this reason that many pollsters (like the aforementioned PPP) prefer party self-ID to registration. Indeed, about 20 states don't even have registration by party, so this kind of data doesn't even exist. And in certain places, like some Southern states, there's still a considerable "Dixiecrat overhang," as I like to call it, whereby lots of voters are registered as Democrats due to tradition, even though they probably would never pull the lever for Team Blue. For instance, almost 55 percent of Kentucky voters are still enrolled as Democrats, despite the state's conservative nature.

It's not clear to me why MassINC likes to use registration figures, particularly since they themselves acknowledge that the overlap with partisan self-identification "is not close to perfect." Indeed, given how fluid party ID can be, it seems like you can miss real movement if you rely on registration instead of self-ID, since it's a lot easier to change your mind about what party you identify with when talking to a pollster versus actually bothering to fill out a new voter registration form. But this is still an interesting and illustrative example of the differences between self identification and voter registration statistics.