An obvious comparison is to ask how Ryan performed compared with Sarah Palin, another running-mate intended to be a game-changer.
As you can see, a major advantage of the Meta-analysis is that it allows high-resolution tracking over time. This kind of precision is hard to achieve even with national polls, probably because there are fewer of them. Here are my thoughts on this comparison:
For Palin, I have indicated key events that seemed significant at the time: her electrifying RNC speech (9/3), the airing of her disastrous interview with Charlie Gibson (9/11), John McCain’s appearance on The View (9/12), the bankruptcy of Lehman Brothers (9/15), and the first Obama-McCain debate (9/26). In each case the arrowhead indicates the expected direction of movement.
As I noted before, any race-shifting event would make itself apparent after a delay of a few days. In 2008 the post-Palin-speech delay appears even longer, probably because state polls were put on hold during the Republican National Convention. I am struck by two points.
First, the post-Ryan shift is almost twice as large as the post-Palin bounce. I find that a little surprising. Then again, the Ryan announcement was preceded by a month of very effective attacks on Romney over his tax returns and Bain Capital involvement. And Palin’s appeal at the time may have been focused on mountain/frontier states, which were largely not in play. It’s hard to say where the natural baseline is for judging the size of these shifts.
Second, McCain peaked on September 15th, 12 days after the Palin speech. If the reversal for Romney on August 22 is the corresponding peak, that would be 11 days, very similar. In both cases, two weeks was long enough to gauge the initial shift. The Ryan bounce might now be declining more steeply.
Next, the Republican National Convention. The convention has only three rather than four days to make an impression because of Tropical Storm Isaac. Luckily, their lineup is shortened by the absence of some major figures, Sarah Palin and George W. Bush. Seems like a good idea to me.
An aside: the Meta-margin bounces look similar when plotted in units of electoral votes:
Over the range shown in these graphs, 1 point of popular vote margin translates to about 12.5 EV, in a linearly proportional manner.