Nate Silver of FiveThirtyEight.com published a short, sidebar analysis in the latest issue of the New Republic showing that in recent decades the candidate with the larger post-convention poll bounce invariably wins the national popular vote. That is, Al Gore out-conventioned George W. Bush, as did the other outright presidential winners. (TNR requires a subscription, but its link to Silver's analysis and accompanying chart is no longer available anyway; however, here is a more technical post from Silver about convention bounces.)

Mark Penn, in the Politico today, echoes Silver's point:

This year, the party that wins the battle of the conventions will likely win the election. In the past 60 years, few presidential candidates have overcome negative poll numbers taken after the conventions. While races have gotten closer and debates have had an effect, nothing in the months between convention and election has swayed the voters' preferences.

With all the ads, debates and spin and counterspin, can it really be that the conventions -- usually dismissed in the modern era as confections, as mere pageants -- matter so much? The evidence seems to be there. This could just be a reflection of how close elections have been in recent decades, and hence one can find any number of so-called pivotal factors.

Still, it is amazing, if a bit daunting, to think that these two, four-day pageants held during the late-August/Labor Day period, when voters are supposed to be tuned out at the beach, watching the Olympics or back-to-school shopping, can matter so much to the final outcome three months later.

(Note: Salon will publish a "conversation" with three experts about the meaning and significance of party conventions later this week. Stay tuned.)