Princeton Election Consortium

A first draft of electoral history. Since 2004

Jul 12: Biden 381 EV (D+6.3% from toss-up), Senate 53 D, 47 R (D+5.6%), House control D+8.0%
Moneyball states: President AK AR IA, Senate MT KS ME, Legislatures KS TX NC

Watching the next wave break

September 12th, 2008, 5:12pm by Sam Wang

Now that it’s Friday afternoon, let’s review where things stand.

Electorally, the race is a near-perfect draw. The question is: what’s next? Here’s my take on what to watch in the coming days.

The national polls. These are interesting, but not because they tell us who would win today. Only the Electoral College tells us that. National polls are worth watching because they are done frequently and therefore reflect movement quickly. This week the record is consistent: McCain holds a steady 2% lead over Obama in popular opinion. The current median margin for September 9-11 (Gallup, Diageo, Rasmussen, DailyKos/R2000) has McCain ahead by +1.0+/-1.3%.

The state polls. As I reported this morning, these are nearly complete. There are a few holes in the data. Watch NM, MT, ND, and NV. In these states the race is close and/or the pre-convention data are sparse. New polls in any of them will move the EV estimator by up to about half their EV strength – most likely toward McCain. Because McCain’s bounce is concentrated in frontier/rural states, I think Montana and North Dakota are likely to be secure for him. My guess based on post-convention polls suggests that the EV estimator is headed toward Obama 264 EV, McCain 274 EV (again, basically a tie given the confidence intervals).

On the flip side, the concentration of McCain’s new support means that his position is not as strong as national polls indicate. As I wrote last night, it looks like he is effectively at a 1-2% disadvantage. If national poll margins narrow to a consistent 1%, I wouldn’t be surprised to see Obama pull ahead in the EV estimator.

Events that move opinion. Here I will take the liberty of being more speculative.

First, an assertion: television drives opinion. Period. Based on a recent Pew Center study, about one-fourth of people get news from the Internet – as you do. Many people use a blend of online and traditional sources. But you are a minority. Put aside the study’s airy-fairy future-babble: 60% of people listed television as their first source of information.

Visually presented information is readily assimilated, as I pointed out in a piece I wrote about false belief formation. When something gets on television, it can move opinion in quantity. Examples include John McCain’s houses gaffe and Sarah Palin’s convention speech.

Two recent televised events may have potentially influenced opinion: John McCain’s appearance on The View, and Sarah Palin’s interview with ABC’s Charlie Gibson. What I’m reading isn’t positive for the McCain campaign. But I didn’t see what it looked like. I’m a stereotypical pointy-headed academic, no television. So I’ll just say that there is the potential for movement in the coming days.

Even if I’m wrong, another statement is highly likely to be true: The first two debates on September 26th and October 7th are key decision points (and the VP debate on October 2nd). If the race remains close they could be deciding. The 2004 campaign provides an instructive example.

Median EV estimator from 2004 race

Median EV estimator from 2004 race

As you can see, the first debate that year was almost game-changing. But not quite.

Tags: 2008 Election

11 Comments so far ↓

  • Independent

    The McCain campaign seems to have profited from your article on “false belief formation.” Sarah Palin’s repetition of the “Thanks, but no thanks” embellishment, McCain’s and several of his campaign operatives’ ludicrous but repeated claim that proximity to Russia made the governor of Alaska an expert on foreign affairs, etc. all point to calculated efforts to exploit “source amnesia.”

  • San Fran Sam

    I am wondering if there is another influence on the polls. Namely, the polls themselves.

    Americans are very low information decision makers. to the extent that the horse race, ie national, polls are constantly being reported, what influence will the trend have on future polls.

    That is, as long as the MSM keeps saying things like “The trend in the polls is towards x” then future polls will further trend toward X and the election results.

  • Bob Rimmer

    The meta-analysis of the data does not reflect trends particularly well. A stable analysis of the intrade data shows a rather dramatic change over the last few days. In polls one is generally asked if the election were held today, who would you vote for. The intrade participants are betting who will win in November. The pollster’s question leaves open the possibility of changing one’s mind. The futures market is betting on the outcome. Nearly $10 million is currently being traded daily. The participants in this market may be mostly Europeans, but the change in the outlook has nevertheless been dramatic.

  • Fred

    1.The conventions often produce a bounce (huge amounts of free, positive publcity)–first a boost in popularity, but then a decline.
    2. I know Palin is felt to have had a huge impact, but my impression is that vice presidential choice usually has very little over all impact.

    Therefore, I’m cautious about givng too much credence to the present McCain spike. I’m curious to see if his numbers decline in the next week or two, as the bouncing ball comes back down (closer to pre convention levels).

  • Richard Wiener

    Since you took the liberty of being more speculative, I will follow your lead. Of course, TV plays a critical role in molding public opinion. But there is something even deeper, a public consciousness fed by all the media (even the internet). What has changed since Palin is the perception that McCain can win. Most people suddenly sense that McCain is on a winning arc, that his campaign outmaneuvered the Democrats and Obama is floundering in response. Maybe the debates will shake things up, as you suggest, but I doubt it. Something big like a McCain or Palin sex, alcohol, or drug scandal seems necessary. Otherwise, I think the race will slither along to a 2004 EV outcome.

  • David

    It’s nauseating sometimes to know CNN decides every election. The same network that thinks global warming should be debated with oil lobbyist, and Larry King should talk about psychics, ghosts and vaccines/autism links every other night. Thank God the NY Times exposed the network propagandizing people with generals from the pentagon earlier this year.

    I’m glad the comments work again. I wanted to ask you some questions before but now I forget what they were.

    If only Ohioans knew about half of what’s reported on Digg or the Daily Kos. Hardcore Republicans would be outraged if they saw past the media filter.

    You should probably add a pointer on your graph now to include the conventions causing the dip in the polls.

  • Heasun Choung

    Sam –

    What’s your take on the “Time-for-Change” predictor model as proposed by Alan Abramowitz?

    Essentially, his argument is that there are three primary factors that are useful predictors for election winners.

    Popularity of Incumbent President
    Economy (GDP growth)
    Number of Terms for Incumbent Party

    The results he comes up with based on hismodel is that the “time-for-change model predicts that Barack Obama will receive 55.1 percent of the major party vote in November vs. 44.9 percent for John McCain.”

  • Frank

    Am I right that the distribution is getting less variant and the average values therefore greater? The modal outcome (280?) of about 9% — is this one particular configuration of states and if so what?

  • Max Neiman

    It’s worth mentioning that with financial markets as sensitive as they are, that something dramatic in the economy could happen at any time. For some reason, folks seem to have become less aware of how weak the economy is. Additional bad news on unemployment might add to the rocky picture. In other words, it’s possible that the public might turn from the Circus Maximus they’re being provided just now — football, hurricanes, and the Palin smokescreen. I also get the sense that the Obama campaign is developing a fairly unprecedented get-out-the-vote apparatus, which could be important too in what seem likely to be a close election. Finally, since this source seems to be so important nowadays, I am getting this gnawing feeling in my gut that Obama might replicate Bush’s feat in 2000 and win in the Electoral College, but lose in the popular vote. That would be delicious.

  • New Hampster

    Great site, and very useful analysis.

    In addition to your caveat that the national polls “are interesting, but not because they tell us who would win today — only the Electoral College tells us that”, a second needs to be added, however.

    The great majority of the polls cited and arithmentically averaged thus far are based on self-reported registered voters, not “likely voters”.

    As the election draws closer, pollsters typically shift their public “horse race” reporting from the former to the latter — that is, to basing their public numbers on (only) that sub-portion of respondents they believe is likely to actually cast votes on Election Day.

    Many items go into helping estimate/define the latter, and pollsters differ in both their estimate of probable turnout and the specific questions they use to try to determine any given respondent’s likelihood of voting.

    For any given poll and pollster, however, there is often a significant difference between the candidate preference levels reported by all registered voters and those among that same poll’s “likely” voters (as estimated by that pollster).

    Often — though not invariably — this “re-basing” shift itself produces numbers more favorable to the GOP. Time will tell if this is the case this year… my point is simply that this is one more element of (current) inexactitude to bear in mind as regards both national and and state polls two months prior to election day.

  • Sam Wang

    Independent: I have come to the conclusion that Republican strategists understand very well how the brain forms beliefs, in a practical sense. So even though it looks as if they are applying our article, I would say instead that we have spelled out some of those strategies in terms of neuroscience for everyone to understand.

    jc: Elections vary from year to year. See the histories of various campaigns over the last few decades, plotted by Charles Franklin at (linked on this site at some point in the last week). However, in general you are right that having the convention second seems like a winning move. There might be an extreme version of that this year: the Democrats had a great convention that was memorable, but these memories were crowded out almost right away by a Palin-fest. I hope we get past this and back to issues.

    Heasun Choung: I think that predictive model is interesting, but basically a quantitative formulation of what people write about Presidential elections. Personally, I think Obama is likely to win in November for reasons like the ones mentioned in that model. However, 55% of the popular vote? Please. if he gets even 53%, I will wash Alan Abramowitz’s car for a month.

    Frank, the distribution seems tighter. This is visible in the history plot as well. The mode is usually the highest peak (which is currently at Obama 276 EV). However, it doesn’t have to be, because each number of EV contains many combinations. The mode can be seen in the right sidebar (click a state to force state assignments, then click it more to force it to the “right” color). Today the mode is 273 EV and contains Obama CO, MT, NM, NH; McCain OH; and all the less surprising outcomes.

    Max, the popular / EV vote split you mention is quite possible – perhaps more possible than it’s been in decades. Considering the recent nastiness of the campaign, it would be better for the win to be non-contentious.

Leave a Comment