Convention Bounce

Assessment

Standards for assessing the bounce differ. While it sometimes is reported among “likely voters,” it is more meaningfully assessed among all registered voters, which is a more stable and more uniformly denned population. And the fullest picture can be drawn not by looking only at change in support for the new nominee, but—offense sometimes being the best defense in politics—at the change in the margin between the candidates, to include any drop in support for the opposing candidate. For example, the 1968 Republican convention did more to reduce Humphrey's support than to bolster Nixon's.

Timing can matter as well; surveys conducted closer to the beginning and end of each convention better isolate the effect. In 2004, Gallup polls figured John Kerry's bounce from a starting point measured 5 [Page 150]days before his convention began and assigned him a net loss of 5 points—its first negative bounce since McGovern's 32 years earlier. Using different timing, ABC News and The Washington Post started with a pre-convention measurement done 4 days later than Gallup's, and found an 8-point bounce in Kerry's favor, much nearer the norm.

Using the change in the margin, among registered voters, the average bounce has been 10 points in Gallup polls from 1968 through 2004 (and, for comparison, a similarly sized bounce of 13 points in ABC News polls from 1992 to 2004). While individual bounces vary, on average they have been consistent across a range of parameters: in Gallup data, 11 points for Democratic candidates (9 points leaving aside Clinton's 1992 bounce), 9 points for Republicans, 8 points for incumbents, 11 points for challengers, 10 points for better-known candidates (incumbent presidents and incumbent or former vice presidents), 10 points for less-known candidates, 12 points after each cycle's first convention, and 9 points after the second convention.

...