They're Called Convention Bounces for a Reason


For those of you new to the dynamics of U.S. presidential elections, the Democratic and Republican Conventions each tend to create a “convention bounce”: an increase in the poll numbers for their candidate and his party.  The size of these bounces varies from year to year and convention to convention, but the reasons for their existence are clear.  While conventions were once where presidential candidates were actual chosen, where party leaders entered the infamous smoke-filled rooms and hashed out the deals that would give someone the majority (or in the case of the Democrats before the 1930s the supermajority) of the delegates necessary to secure the nomination, since the 1970s, there has been little real drama about who will emerge as the parties’ nominees once the conventions role around. As a result, conventions have come to play a rather different function: they are vast, multi-day advertisements for a party and a candidate.  And though the death, or at least decline, of the convention has been predicted since they became a form of infotainment, the institution seems still to be going strong.  It’s not surprising that the convention-as-multimillion-dollar-advertisement tends to garner an increase in market share for the advertised product.  But this increase tends to be temporary.  As the campaign returns to its usual give-and-take between the two candidates, the convention bounce dissipates.  And for the party that has its convention first, the other party’s convention usually ends whatever remains of the first party’s convention bounce.

This year, for the first time ever, the conventions occurred in consecutive weeks, with the Democrats going during the last week of August and the Republicans in the first week of September. In addition to the usual speculation about convention bounces, then, this year there was the added speculation about how having one convention immediately following the other would affect the phenomenon. On August 15, well before either the Democratic or Republican Convention had taken place, Nate Silver, one of the most astute poll analysts working today*, made some guesses about what we might expect from convention bounces in this year’s presidential race.  Silver suggested that if you simply ran the effects of two average convention bounces, one after the other, you’d get a graph that looked something like this:

Silver’s guess proved to be remarkably prescient, especially if you factor in Obama’s slim, pre-convention lead.  About the only way in which polling has deviated from this pattern is that McCain’s convention bounce has deteriorated somewhat more quickly than Silver predicted.

What I find interesting about this is that although the pattern in polling over the last several weeks has followed Silver’s expectations, which were in turn based on typical convention bounces in the past, those weeks have been experienced as anything but typical. McCain’s surprise choice of Sarah Palin had appeared to vault him into the lead and fundamentally change the dynamic of the campaign. Then, starting the week before last, Palin’s numbers seemed to fall in the wake of the public learning a bit more about her. And last week, the extraordinary crisis in the financial sector—which had been brewing for months but coalesced suddenly and in ways that could not have been predicted—played squarely into Obama’s strengths and McCain’s weaknesses and apparently vaulted Obama back into the lead. 

And so we find ourselves about where we were before the conventions, with Obama holding a small, but significant, lead that reflects a number of structural advantages that he and his party have had all year.  But with only forty-four days to go until electionday, and the conventions behind us, McCain’s opportunities to overcome his disadvantages are growing slimmer.

* Nate Silver’s blog, (named after the total number of electoral votes), has become essential reading during this campaign season.  Silver was previously best known as an analyzer of baseball statistics and Managing Partner of the Baseball Prospectus, which is in effect a baseball statistics think-tank.  Other worthwhile analyses of polling data can be found at and at Prof. Sam Wang’s Princeton Election Consortium




Thanks for this post! I was wondering if there were any statistics web pages you’d recommend. The German media tends to report all shifts in the popular vote as if they mattered… So I was looking for good statistics on the expected distribution in the electoral college (beyond the NYT map that I like for being interactive that, unfortunately, is not updated very frequently). seems like an awesome resource.

Talking of the media: Watching coverage around the time of the convention, one of the things that baffled me most was how much the media expected the convention bounce to be nothing but a bounce in the first place, an expected and calculated effect. Yet, at the same time, they reported it as breaking news and as if they didn’t know that the excitement (and even some of the Palin-enthusiasm) was ‘just’ the expected bounce. And in a way, I think this is a very real dilemma for the media in this election. Would you agree that, more than in previous elections (or, let’s say, more so than before the 2000 election ?), the media seems to be very much aware of the narrative dynamics it is caught up in—but there seems to be no way to move beyond this…

Great comment, Sebastian.  I think you’re absolutely right about the state of U.S. media coverage.

As I indicated in today’s post on debate season, the media has gotten ever more self-conscious and sophisticated (or at least pseudo-sophisticated) in its coverage of our political campaigns. A related trend is the continuing growth of “horse-race” coverage (focusing on who’s up and who’s down and, in very conventional ways, why) at the expense of acting as a kind of protector of a more robustly conceived public interest (in either the Woodward and Bernstein adversarial model or the more old-fashioned defense of what was once called the Establishment…the latter style of journalism is still represented by someone like David Broder).

Of course along with this sophistication came ever more naked attempts by campaigns to manipulate coverage, attempts which were then eventually covered by the press even as they were being manipulated.  After a couple decades of this, a pretty sophisticated metadiscourse about the whole mess has developed.  Terms that start in communications theory, political science, or political campaigns—spin, framing, Overton Window, and so forth—slowly work their way into the national political discourse, a process that has sped up exponientially as the twenty-four hour network newscycle interacts with the Internet’s native logophilia.  This year the word “narrative” seems to have suddenly exploded into mainstream media discourse as “spin” did in the late 1980s. 

Despite this metadiscourse, the media is still focused on reporting the horserace in news-cycle-by-news-cycle ways. Thus they can be simultaneously sophisticated about expected convention bounces and enthusiastically surprised by the unexpected Palin pick, even though the exact same poll numbers are taken as indications of both phenomena.

And despite the media’s newfound postmodernism, as the case of Ron Fournier suggests, the manipulation of the media by the campaigns remains a fairly old-fashioned business in many ways.