Here’s a new report by Third Way. They find that in 5 of 8 battleground states that register voters by political party, the number of registered independents is up. In 7 of 8 states the number of registered Democrats is down. The GOP is down in 6 of those 8. They then write:
Beyond these battleground states, national surveys such as the American National Election Studies and Pew show a steady increase in Independent self-identification throughout the United States. According to Pew, between 2000 and 2011, both the Democratic and Republican parties lost members, and the number of self-identified Independents increased by 8%. In 2000, 33% of the electorate identified as a Democrat, 28% as a Republican, and 29% as an Independent. By 2011 only 32% identified as a Democrat, 25% as a Republican, and 37% as an Independent. Democratic and Republican losses were mirrored by gains in Independents…
No acknowledgment of the fact that most of them lean toward a party and tend to vote loyally for that party. Or that presidential candidates routinely lose independents but win elections (at least the popular vote). See Jimmy Carter, Al Gore in 2000, and George W. Bush in 2004. (The report mentions Gore but not Carter or Bush.)
Okay, let me stop harping on that and make a more constructive point: despite the million stories you’ll read between now and November 2012, it’s not very instructive to consider how a candidate is performing among any particular group of voters. Taken in isolation, this means nothing. The candidate could offset those losses with gains among another group or groups.
Or let me give an even more plausible scenario. If a candidate is losing votes among some group relative to a past election, chances are they are losing votes among many groups. Just as states often exhibit a uniform swing in response to the structural forces that affect election outcomes, so do groups of voters. This is why I wrote a few weeks ago, apropos of Jewish voters:
What’s happened to Obama is not a “Jewish problem.” It’s an “economy problem.”
Or consider the shifts in various groups from 2008 to 2010 (the graph is something I pulled from the New York Times months ago but I can’t find a link):
Democrats lost votes among men, women, blacks, whites, Catholics, Protestants, young voters, old voters, etc. Sure, the shifts within these groups were not all identical, but 2010 is much more about the similar direction of the shift, not the differences among groups.
For this reason, I’d be a lot less interested in where any particular group stands—or in identifying the groups that are allegedly the key to victory—and much more interested in the factors that shape where all groups stand.
{ 2 comments… read them below or add one }
All strong (and repetitive) observations. Unfortunately, we nearly always measure party identification and vote choice simultaneously. In most cases, we don’t know how many independents are temporarily leaners once they have chosen a presidential candidate or otherwise show weak commitment to a candidate over the course of a campaign.
Bartels, “Partisanship and Voting Behavior, 1952-1996,” AJPS (2000):
“An attractive alternative approach is to ‘purge’ current partisanship of short-term influences by using lagged partisanship as an instrumental variable. . . . Instrumental variables (IV) estimates of the impact of partisanship on presidential voting behavior in 1960, 1976, and 1992 are presented in the third row of each panel of Table 4, and the corresponding estimates for congressional voting behavior are presented in the third row of each panel of Table 5. By comparison with the original results presented in the first row of each panel, the instrumental variables results imply slightly lower levels of partisan voting in presidential elections (by 8 percent in 1960, 2 percent in 1976, and 12 percent in 1992) and somewhat higher levels of partisan voting in congressinoal elections (by 9 percent in 1960, 5 percent in 1976, and 27 percent in 1992). These results suggest, quite sensibly, that for presidential voting the upward bias due to short-term influences exceeds the downward bias due to measurement error, whereas for congressional voting the bias due to short-term influences is smaller than the bias due to measurement error.”
In the analyses of presidential voting, the ordinary probit coefficients for “leaners” are, on average, 21% larger than the corresponding coefficients for “weak” partisans and 39% smaller than the corresponding coefficients for “strong partisans.” The IV estimates for “leaners” are, on average, 65% larger than the corresponding coefficients for “weak” partisans and 35% smaller than the corresponding coefficients for “strong” partisans. So, if anything, taking account of the endogeneity of contemporaneous partisanship reinforces John’s point.