Larry Sabato and his team studied party convention bounces going back about a half century and found the average bounce in the polls is about five points.
Key finding: “The actual standings of the candidates after both conventions are sometimes amazingly predictive of the November results, and at other times are terribly misleading. So wouldn’t you know it? We just can’t rely on bounces to tell us much beyond whether the electorate is ‘dug in’ and resistant to switching sides. In 11 of the 24 presidential candidacies since 1964 noted in the charts above, candidates finished within three percentage points of their post-convention polling. Statistically, this makes the bounce meaningless in a predictive sense, because about the same number of candidates’ standings did change significantly from the end of their convention to November.”