Campaign events vs. the fundamentals

We're at a historic moment -- Barack Obama's estimated lead in the national polls is over seven percentage points:

There are two ways to interpret what's happened. Most journalists will soon converge on some narrative where John McCain has lost ground due to some combination of the financial crisis, the incoherence of Sarah Palin's TV interviews, and McCain's allegedly "small and angry" debate performance.

For instance, The Atlantic's James Fallows, who previously made the unsupported claim that "moments" from televised general-election debates have "figured in the ultimate outcome" in the presidential elections of 1960, 1976, 1980, 1984, 1988, 1992, 2000, and 2004, now refers to John McCain's "sourness and anger" during the first debate as a "moment" that '"mattered'":

"Everyone knows," based on a long string of past episodes, that some unintentional flash of character revelation usually turns out to be the memorable aspect of a presidential debate. Eg: Nixon looking furtive and sweaty in 1960, Ford momentarily seemed befuddled in 1976, Dukakis seeming bloodless in 1988, etc. All these moments "mattered" because they crystallized a feeling that, in retrospect, people knew they'd "always had" about the candidate.

In the days since the first Obama-McCain debate, it's become ever clearer that John McCain's sourness and anger are the traits unintentionally revealed in the debate and now working against him... Thanks to McCain's hostile refusal to engage Obama as a human equal face-to-face at the debate, the image he is cementing is that of a seething older man. Like Bob Dole in 1996, with less of a gift for one-liners.

It all fits into a pattern in retrospect -- but I don't know a single "expert" who predicted that avoiding eye contact would be the enduring image of the first debate.

The key phrase is "It all fits into a pattern of retrospect," which is precisely the point -- humans are quite good at constructing post hoc narratives to explain observed events. Princeton's Larry Bartels recently recounted an amusing anecdote along these lines:

I was very struck when I learned -- many of you probably have seen, after each recent election, immediately after the election, Newsweek comes out with a big cover package on why, fill in the blank, won the election. And in 2004, they actually came out with a book that included a lot of analysis of why it was that Bush won the election. But before the election, they actually sent out an advertisement that had two books side by side; one was why Bush won the election and the other was why Kerry won the election. And given the times of producing these things, they actually had to produce most of the package, explaining to the readers of Newsweek the following week why it was that Kerry won the election.

Now, I didn’t read that issue, but I’m pretty sure that if I had read that issue, the narrative of how it was that Kerry had won the election would have been about as convincing as the narrative of how it was that Bush won the election.

So after the fact, it’s really easy to come up with explanations based on particular events, but to know whether those were the events that actually drove the usually pretty small deviations from the underlying fundamentals that we observe in a particular campaign is quite difficult.

As Bartels's answer suggests, I think a better explanation is that the underlying political fundamentals are reasserting themselves -- the dynamics of this year are simply remarkably unfavorable to Republicans. This account is consistent with Andrew Gelman and Gary King's research showing how voters tend to converge to predicted levels of support for the presidential candidates over the course of the campaign.

Looking back, there's a similar divide over 2004, which is frequently framed as a story in which Karl Rove's tactical mastery and the Swift Boat ads triumph over John Kerry and his ineffective campaign tactics. The problem for Kerry, however, was that he was always fighting uphill. Political science models of the election published in the journal PS before the election had a median forecast of 53.8% for President Bush's two-party vote:

Now, it is true that Bush's numbers bounced around over the course of the campaign as various events occurred, as Charles Franklin's plot below illustrates:

We can certainly tell stories about those events (such as the Swift Boat ads). However, the ultimate outcome was relatively close to the projection -- Bush ended up winning as all the models projected (except for one projected statistical tie of 49.9%) and received 51.2% of the two-party vote.

This year, the comparable set of forecasts in PS, which was just published, shows a median two-party vote forecast for John McCain of 48%, which corresponds to 52% for Obama. Only one model projects that McCain will win:

If you include the projections from Ray Fair and Douglas Hibbs of Obama receiving 51.5% and 52% of the two-party vote, respectively, that I've previously highlighted, the median forecast remains Obama 52%, McCain 48%.

Obama's lead now is consistent with those projections (he's currently at 53.9% of the two-party vote in the Pollster.com estimate above). In fact, despite my previous warnings that Obama may underperform due to the influence of race, he is currently in a very strong position -- the fundamentals favor him and he has a lead with approximately one month to go. Despite media hype about "October surprises," these these plots of recent competitive presidential campaigns from James Stimson's Tides of Consent illustrate that there are rarely dramatic shifts in the polls this late in the game:

That's why I expect John McCain to start taking more risks in the next couple of weeks, particularly at the next debate. Expect to start hearing a lot about Tony Rezko, Jeremiah Wright, and William Ayres. They're the three cards that McCain has yet to play.

Update 10/5 10:40 PM: Matthew Yglesias relates an anecdote that mirrors the one from Bartels above:

My mother worked in the pre-Photoshop version of Newsweek’s art department so I saw as a kid their “Dukakis Wins!” complete with a banner teasing their account of how he did it and why the prognosticators were all wrong. As Bartels says, most of that account would have to have been written in advance.

Update 10/13 10:37 PM: Kevin Drum makes a similar point.

Comments

[tristero]

Pardon my blithering stupidity here but I do have to ask some really dumb questions.

At least twice you mention that the fundamentals favor Obama. But in re-reading your post, I still don't have a clear idea of what those fundamentals are. We talking economics? Or values? The war?

If so, why wouldn't dramatic events in each category affect the campaigns?

And why wouldn't a cumulative affect make a difference?

If they don't, why bother with elections at all? Just look at these PS polls, determine the winner and save everyone a heap of trouble.

[David]

tristero, from other posts of Brendan's, I believe that "fundamentals" refers to economic conditions as measured by certain indexes (although there may have been some additional items included in that term.)

Note that the economy is "fundamental" in the sense that it reflects the real world, as compared with campaign ads, debate fluffs, media coverage, etc. However, I believe that the term "fundamentals" does not include certain other items with important external reality, such as the crime rate, wars, national security, pollution levels, perceived moral climate, availability of abortions, etc. Even the national mood has a kind of external reality, even though it can't be easily defined. As I recall, a part of JFK's popularity came from the pride he brought to the country.

In my cynical opinion, academic political scientists are apt to define "fundamentals" as consisting only of items with numerical measurements. Then they can apply the tools of statistics to their data and crank out publishable papers. They don't want to focus on Gerald Ford's debate fluff, Al Gore's sighs or the media infatuation with Obama, because these things cannot be included in an Analysis of Variance. This sort of item does fulfill the need of pundits who write regular columns. And, practical politicians no doubt have their own slant on what the key factors are in winning an election.

[Dan Ryan]

If McCain is going to go the Rezko/Wright/Ayers route, doesn't he need to say something new about them?

Right now, the narrative is that he is flailing and we are all trying to pick apart the myriad reasons. If he can't say something new about Rezko/Wright/Ayers, it's more likely to feed the flailing narrative than change the conversation.

[MartyB]

Dan -

I would guess that most people know little about Obama's associations with Rezko and Ayers (my guess is they probably heard something about Wright since that got some media play last spring), since the media has downplayed and almost completely ignored these two relationships significantly. So even old information might appear "new" to many people.

As an aside, I doubt if any Republican candidate who had gone to Falwell's church for 20 years without complaining about some of his outrageous comments, or sat on a board with the unabomber and then said he was "just some guy for my neighborhood", or received some unusual help from an indicted racketeer in buying his house, would have treated this daintily by the media.

Even having said that, I wonder if McCain bringing these associations up will have much impact. The “undecided” people I’ve talked to seem to really want to vote for Obama – they like the way he talks, they think it would be good for the country to elect a black man (which by the way, is how I believe race has and will continue to effect this election in the most profound way – as opposed to the standard trope about bringing out the racist vote) – and they seem to be looking for reasons to vote for him. I don’t see that a lot of focus on Ayers, Rezko or Wright will have enough impact to overcome this phenomenon.

Another aside: I have been reading the book _Blink_ recently and been wondering if we aren’t witnessing a version of the “Harding Effect” – that Warren G Harding looked presidential, and was selected by his party and elected primarily on that basis, but turned out to be one of our worst presidents.

I guess only time will tell.

Maybe Obama will step up and be a great president, but there really isn't anything in his history that seems to indicate this. He gives good speeches. What else is there to him that isn't just wishful thinking?

[Hemingway]

Marty B,

I don't know what distracted your attention throughout the month that the Rev. Wright soap opera was the hottest item in the media meat grinder, whipped into a lather by the likes of Rush and Bill O'. In fact, because the loquacious preacher would not STFU, Obama left his church.

As any member of a religious institution knows, the viewpoints, even doctrines, of its leadership are not followed blindly by all. The obvious example: no matter what their infallible Pope says,how many Catholics would never use contraception?

On the other hand, we actually have video footage of Palin actively seeking her preacher's protection against witchcraft, so that obviates any plausible denial from the McCain Campaign. In my book, her religious beliefs don't matter a scintilla, but at least Palin's belief in witchcraft is not a case of spurious guilt by association that you feel the news media should be dredging up to clobber Obama with again.

[Mark]

Tristero,
The concept of "fundamentals" refers to political scientists' predilection for parsimony - trying to explain a lot with a little. Many of these predictions are labeled "scientific" because they take the same factors each time around and use them to make decent predictions. The factors labeled "fundamentals" typically include aggregate economic and national security data (e.g. GDP growth, unemployment, war deaths, etc.) Each model uses slightly different measures with different weights for those measures, but they often look at similar data, which we've come to call the fundamentals. Brendan is trying to point us, for better or worse, toward a more scientific perspective on elections. Rather than ad-hoc and post-hoc narratives, I think he wants the media to focus on more predictive and reliable factors. This doesn't mean we lose our individual will; it just says we as Americans arguably have some regular behaviors at the aggregate level.

[kandis]

Brendan,

You are my mainstay, helping me understand the macro view of these political events. I will continue to follow your writings and anticipate many, many years of enjoying the intellectual fruit you share with us.

I have a micro piece that your soon to be published study inspired:
As I watched the debate between Palin and Biden, I was suprised by how very much Palins’s attitude that she didn’t have to follow the rules disgusted me. “ … I may not answer the questions the way that either the moderator or you want to hear but I’m going to talk straight to the American People and let them know my track record…”. She really hit a nerve. I could not listen to her with an objective mind because she had declared that she wasn’t going to play fair. Personally, I have had quite enough of leaders that assume that the rules apply to others, but not them.

Kandis

[FermionTheClown]

To respond somewhat whimsically to tristero at 11:55AM 10/3, there's a famous line in Asimov's famous Foundation series which goes something like this: "the behavior of a single man is predictable by no known science; the behavior of quadrillions is completely predictable."

Classic early sci-fi (1950), it's a painless read that makes intuitive the program of applying quantitative modeling to human behavior.

uh, but fiction.

[Jinchi]

Tristero, Brendan has a lot of posts about "the fundamentals" in American political elections, a basic primer is here (www brendan-nyhan.com/blog/2008/02/projecting-the.html).

These are typically 2 variable models (the bread and peace model uses a combination of the growth of per-capita disposable income and the number of U.S. military fatalities). I believe these models suffer mainly from a lack of data (there are only 14 elections in the bread and peace model) so there's a large amount of error in the final prediction (the elections of 2000 and 1996 both fall 5% off the predicted line, while 1952 and 1968 dominate the "peace" part of the equation).

Still it's a better baseline to start from than nothing and certainly better than the analysis of sighs, eye-rolls and the angry demeanor of candidates at the debates.

[MartyB]

Hemingway -

Apparently you mis-read my comments as I said "my guess is they probably heard something about Wright since that got some media play last spring", so it doesn't seem your comments about my missing the coverage are relevant.

No need on your part to issue a "nevermind" - it's understood. ;-)

MartyB

[ChopLogic]

Explain to me where I go wrong.

In Part A, the bulk of the blog, Nyhan's message is that campaign events don't really matter that much, they are only made to seem to in retrospect. What matters is the "fundamentals." which tend toward a certain vote share and the effect of which will determine the outcome, despite the vagaries of campaign events.

In Part B, the tail end, Nyhan becomes incoherent by suggesting that after campaign events--that is, ads and speeches invoking fear and disgust over Obama's association with Ayers, Wright, Rezko--may well work to turn the election in McCain's direction. Or at least they're his best chance. Anyway, there the sort of thing that CAN determine the outcome of an election.

So why am I wrong? Why doesn't little B take back, contradict, big A?

And, please relate the apparent dismissal of the efficacy of campaign events to Sam Wang's charts purporting to show an immediate and large impact of certain campaign events--e.g., of the celebrity ad, of the houses gaffe.

[tristero]

Thank you for the help, people, in understanding Brendan's post.

[Jinchi]

So why am I wrong? Why doesn't little B take back, contradict, big A?

At the risk of putting words in Brendan's mouth, I think he's saying that the fundamentals favor Obama, but don't give him the landslide victory that many think 8 years of Bush fatigue should provide. (52%-48% instead of say 60%-40%). McCain can't change the fundamentals, but he has a hope of beating the prediction by 2% (a McCain win of 50.1%-49.9% would still fall within the prediction's margin of error). But he has to get the electorate to focus on something else to do it.

[Sam Wang]

Greetings - Sam Wang here. Thanks, Choplogic, for the mention of my history plot, which shows pretty clearly the effects of individual events.

I have been wondering about this tension between the two ideas of (a) "fundamental" economic and political conditions having a strong influence on the race's outcome, and (b) the role of campaign events.

Fundamental factors might determine a set point for the election, and specific events then affect short-term movements. For example, last year I thought that any Democratic candidate would have about a 5-point popular-vote advantage over any Republican candidate. This feeling was not quantitatively derived, but it was broadly consistent with the models that political scientists use.

One way to think about Fallows's events is that they can either drive the race toward its natural equilibrium, or be brilliant acts of stagecraft that can, on their own, move the race.

In the natural-equilibrium category are economic events (long-term or short-term, such as the Lehmann collapse) that set the stage, after which it takes an actual debate to get people to pay attention and therefore move opinion.

In the stagecraft category is the addition of Sarah Palin, which gave McCain a huge temporary boost, but eventually is not seen as playing to "fundamentals." Perhaps she could have been originally seen as being related to Bush's low approval (maverick, change, etc.) but that has dissipated.

Getting back to my history plot, it looks to me like a quantity that tends naturally toward Obama 340-350 EV, McCain 190-200 EV. We'll see when the campaign ends.

Putting this together, my opinion is that most events tend to move the race toward a pre-existing natural equilibrium, but that an inspired move can have some temporary additional effect.