Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

State poll update nearly complete – and a surprise

September 12th, 2008, 6:55am by Sam Wang

The top-line estimate above is nearly current. Since the convention, of the remaining 18 battleground states we have at least 3 polls for PA, MI, CO, VA, OH, FL, NC, and AK*. Nearly all the EV swing would come from these states. We are at a near-tie in the Electoral College.

There is something remarkable going on. The Electoral College has swung significantly less than popular opinion would indicate. In fact, new polls will not push it much further in McCain’s direction. Preliminary calculations indicate that he is at a relative disadvantage of about 1-2% in popular votes. Here is a capsule summary.

We still only have 2 polls from GA, 1 poll each from WI, NM, MT, NH, NV, and MO, and no polls from IA. Based on what we have so far, a full 3-poll median would shift NM, MT, ND, and NV substantially toward McCain.

It is possible to see where the EV estimator is headed. Running the Meta-Analysis using only post-convention polls in states where they are available (and pre-convention polls to fill in the rest) gives a result of Obama 264 EV, McCain 274 EV, with a Popular Meta-Margin of McCain +0.4%. The Obama 95% confidence interval is [229, 300] EV. It’s basically a tie.

McCain’s national-poll margin is currently 2.0+/-0.8%. The discrepancy between this and the Meta-Margin indicates that his popular support is distributed such that he is at a 1-2% disadvantage, viewed through the Electoral College. In other words, in an election held today the likeliest outcome is that he would win the popular vote by several million votes, but barely eke out an Electoral College win. What accounts for this lopsidedness?

I ascribe this to the reason for his bounce: the addition of a base-mobilizing VP pick, Sarah Palin. As pointed out yesterday, McCain has scored large gains in some states (most noticeably Alaska, Idaho, Montana, North Dakota, and Wyoming) but much less in critical states such as Colorado and Pennsylvania. Look at the pattern here: frontier-type states. I offer the opinion that Palin strikes some chord in those states that McCain does not. It could be rural and small-town voters, especially fundamentalists, who are numerous there.

Needless to say, I am now retiring the national-margin adjustment. It’s been completely overtaken by events.

*Only one Alaska poll is post-convention, but all three are after the addition of Palin to the ticket. They show a tremendous bounce for McCain, so this state can be regarded as updated – and completely safe for McCain.

Tags: 2008 Election

6 Comments so far ↓

  • vader

    IMHO, Palin will affect the weak GOP partisan, weak Dem partisan and the undecided voter, negatively. Her press interview was not that great, she is on the cover of the Inquirer with descriptions of drug use and promiscuity in her family and political scandals to fight. Not exactly a good thing.

    It would appear, based on early evidence that McCain motivated the voter that would have voted for him anyway while losing the marginal voter.

  • Independent

    I agree that there seems to be more base rallying in red states than converting undecideds in battleground states that is going on. With the election poised on a knife’s edge as of now, the deciding factors will be (1) the debates, (2) ground operations on election day, and (3) any vestiges of Bradley-Wilder that may be still at play.

  • Fred

    I think it’s great that you’re leaving out the nations margin adjustments. Your research of the state polls from 2004 was convincing to me. ** One should use the pollsters’ data, and not make guesses (no matter how “educated”) about how to correct them.

    Comparison of predictions with outcomes)

  • Nicholas J. Alcock

    J McC thought he needed to shore up his base so he found an Alaskan governor as is running mate. As your analysis indicates this has made little EV dividends. In perhaps, in key respects the Dems must be content that the Reps have reaped their gains in safe states not swing ones? Again, to a comment on the meta-margin site is it possible for the Dems to have more winning EV states i.e. Virginia, Ohio, Nevada, Colorado and Florida without the meta-margin changing? Now, I am not a statistician but my guess is that as the more states are in play in the election the better Obama’s probability is of winning but the meta-margin is likely to increase?. Trying to widen the statewide race is statistically sensible, is it not?

    A few day ago, you argued the Reps spent more advertising monies in tight states whilst the Dems spent more advertising monies across more states. The logic of your argument being the Dems expected a big victory whilst the Reps expected a close one. May I suggest an alternative hypothesis, The Dems have not fought a truly national election since Jimmy Carter in 1976, given the higher registration of Afro-American and younger voters the Dem campaign were kite flying i.e. trying to fing out how much traction the obama had. As we know, the Obama campaign is more flush with monies than the Kerry campaign so this strategy is pretty much riskless?

  • Paul

    Your analysis of the data is, as always, fascinating. But your analsyis shows, I think, that polls are much more useful in providing a an explanation for what has happened than they are predictive. Or, to put it another way, your analysis reminds us that polls are predictive only to the extent that they measure phenomena that they have measured in the past.

    Clearly, the selection of Palin ignited a great deal of interest, and with it the movement of some voters to McCain. But what the polls cannot tell us is how voters will react as the novelty of Palin wears off and they compare what they initially thought about Palin to what they will come to know in the future.

    The changing poll numbers tell us only that Palin has had a strong influence on voters’ preferences right now. But it appears that the future strength of that influence is impossible at this point to predict. This is in part because polling has never measured the affect of a figure similar to Palin.

    I suspect that Palin’s freshness and projection of “small town values” is contending with the voters desire for “change” and competence in government. I have no idea how this will turn out, though, because I’ve never seen anything like this election.

    Thanks again for your terrific work.

  • Sam Wang

    Independent, my gut feeling is that the Bradley/Wilder effect is already contained in phone surveys, some of which are automated. It’s only been observed in exit polls, which involve talking about one’s beliefs to a live person.

    Don’t forget about the question of whether people with cell phones but no landlines are undersampled in surveys. The Pew Center survey on this subject is the best source of information on this. It merits analysis.

Leave a Comment