If this week's convention gives Barack Obama a "bounce," how will we know?
That question may seem silly, given the daily tracking polls and the quickie, one-day "flash" polls conducted to gauge reactions to Obama's selection of Joseph Biden. Never mind the challenges of obtaining a representative sampling of the electorate on a single summer afternoon, pollsters will no doubt pump out all sorts of data this week.
The question of how pollsters will measure convention bounces this year is worth asking because having back-to-back Democratic and Republican gatherings will make it nearly impossible to take measurements of any Obama bounce that are comparable to those from past presidential campaigns.
We have seen a number of historical analyses this week looking back at voter preferences immediately before and after conventions since the 1960s. (The best are from Gallup's Jeffrey Jones, ABC's Gary Langer and University of Wisconsin-Milwaukee political scientist Tom Holbrook.) However, all of these analyses are based on multiday polls conducted entirely before or entirely after each convention.
This year, the two conventions are in consecutive weeks, with a three-day holiday weekend in between. The most important and widely watched event of the Democratic National Convention, Obama's acceptance speech, will conclude late Thursday evening and is widely expected to be followed the next morning, reportedly, by John McCain's announcement of his running mate. In that case, pollsters will have no way to conduct even a one-night poll to gauge the Democratic bounce before the McCain pick the next day, much less the kinds of multinight surveys that measured past convention bounces.
Even if we choose to overlook the impact of the McCain VP announcement, Labor Day weekend presents its own challenges. As I noted last week, CBS News' survey director, Kathy Frankovic, warned that with many people "away from their homes, heading back from summer vacation, or preparing their children for the start of the school year," the holiday weekend will provide "more than the usual number of people who don't respond or can't respond" to pollsters.
Still, if history is a guide, the next two weeks could be crucial for both campaigns. Over the past 25 years, conventions have often marked critical turning points in the presidential horse race. Charts of all available public polls from each year show the convention period to be the decisive and persistent turning points in the elections of George H.W. Bush in 1988, Bill Clinton in 1992, and to a lesser extent George W. Bush in 2004. And although Jimmy Carter in 1980 and Al Gore in 2000 ultimately lost, each enjoyed significant bounces that persisted long after each convention.
On the other hand, many recent conventions produced poll bumps that were fleeting. Ronald Reagan in 1980, Walter Mondale in 1984, Robert Dole in 1996, and George W. Bush in 2000 had bounces that faded away by their opponent's convention, if not sooner.
In reviewing this history, many analysts are tempted to model the past in ways that might help us predict or evaluate the likely course of this year's contest. But searching for consistent patterns in the historical data can be frustrating. As Holbrook points out, "The magnitude of the convention bump is not a great predictor of election outcomes." We all would do well to remember that in looking for patterns in the convention bounces since 1968, we are looking at a sample size of just 10 elections.
So as the conventions progress and the pollsters swamp us with new data, we are probably better off looking beyond the bounce to more specific measures of what voters glean from these party gatherings and how they evaluate that information than the daily twitches of the tracking polls.