Jim Geraghty has been wondering how John McCain could be gaining among independents and still be losing to Barack Obama overall in various national polls. The popular conventional wisdom is that the current crop of polls are oversampling Democrats. Because he specifically asked for the oddly-missing party-identity numbers from the most-recent Los Angeles Times/Bloomberg poll, which has Obama up 46%-44% among registered voters now compared to him up 45%-43% among registered voters in August (conducted before either running mate was made known), despite a swing from a 46%-35% Obama lead among “independents” in August to a 49%-34% McCain lead among “independents” in September, I’ll focus on that. I will ignore the “likely voter” component, where Obama leads 49%-45%, for now partly because I need to compare apples to apples (the August poll did not include “likely voters”), and partly because there is not a partisan breakdown among “likely voters”.
Before I get to the party identification explanation, I must note that Obama’s support among Democrats jumped from 78% to 87%. While some of this came from “McCain-ocrats” (McCain’s support among Democrats dipped from 9% to 5%), most of the gain came from the undecideds, as that dropped from 11% to 5%. In any case, I guess the PUMAs are coming home to roost.
Similarily, McCain has also strengthened his support among Republicans, albeit slightly. That went from 90% to 91%, and appears to have come at the expense of Bob Barr. More-pointedly, the percentage of “Obama-cans” remained unchanged at 6%.
Now, to the math. In order to do a comparison between the August and September polls, one has to know the party-identity internals from the August poll. Fortunately, those numbers are part of the August release, and were 34% Democratic, 29% Republican, 29% Independent, 4% Other Party, and an unmentioned 4% “refused to answer”. Since the remainder of the internals grouped the Independents, Other Party’ers, and “refused to answer”‘s together under the “independent” label, that becomes 37%.
I must point out that the 5-point advantage the Democrats had in the August poll is greater than the biggest of recent election-day splits (4 points in favor of the Democrats in 1996 and 2000, which actually beat the 3-point advantage they had in their historic 2006 election).
This is borne out, within rounding errors, by putting the results of a given early-in-the-poll question with three (or more) distinct answers into equation form (specifically, 3 equations) and solving for the three variables of the Democratic, Republican, and “independent” portions of support for each answer.
Since there are several questions in the September poll that meet those requirements, one can average the rounding errors out. I specifically solved the equations for questions 2, 9 (and the composite “ticket splitter” table), 13, and 15, and came up with an average of 34% Democrats, 43% “independents” and 23% Republicans in the September poll.
While there isn’t a massive increase in the percentage of Democrats included, there are a couple of things of note:
– The split between Democrats and Republicans (+11 D) matches the inflated numbers in other recent polls.
– The 21% decrease in the percentage of Republicans belies the conventional conservative wisdom that the selection of Sarah Palin energized the Republican base, at least on a statistical level. I can cite piece of circumstantial evidence (an incredible number of conservative bloggers jumping on the Straight Talk Express) after piece of circumstantial evidence (60,000 for Palin in Florida) after piece of circumstantial evidence (McCain/Palin outdrawing Obama in Green Bay) why this decrease is a bunch of macaca, but regular readers of this place already know why the decrease is a bunch of macaca.
For grins, let’s apply the August partisan split to the September poll, and the September poll to the August split. Had the August partisan split been applied to the September poll, McCain would be up 46%-44%. Had the September partisan split been applied to August, Obama would have been up 48%-39%.
I’ll briefly discuss the other oddity from the Times poll; the 4-point Obama lead among “likely” voters. There is one question that is somewhat relevant; number 3, or the “are you sure?” question. Again, there is not a specific breakdown of “likely” voters, but neither the partisan split among registered voters (91% of Democrats are sure of their choice, 88% of Republicans are sure of theirs) nor the split between the tickets (85% of Obama/Biden supporters are sure, 84% of McCain/Palin supporters are sure) fully-explains why Obama has that increased difference among “likely” voters.