At a political science conference over the weekend, I saw a presentation by Richard Johnston and Emily Thorson of the University of Pennsylvania. Drawing on the 2008 National Annenberg Election Study, they made the three graphs above.
Graph #1 is the poll standing of the two candidates. The two vertical lines demarcate important moments where McCain’s poll standing dropped.
Graph #2 is the average assessment of the economy. Assessments of the economy were always negative and became more negative as the campaign wore on. But the drops in economic assessments don’t really correspond to the drops in McCain’s poll standing. The first drop in poll standing pre-dates the drop in economic assessments by a few days. The second drop in poll standing comes 8 days after another drop in economic assessments.
Graph #3 is the average favorability toward the presidential and vice-presidential candidates. The trends in feelings toward Palin match McCain’s poll standing almost exactly. The “drops” occur at the same time. It’s eerie.
Of course, Johnston and Thorson describe this analysis as suggestive, and, indeed, treat Palin’s importance as a puzzle — if only because vice-presidential nominees usually have so little impact on the election’s outcome. They write:
bq. Judgment on her was incontestably important. The correspondence between dynamics in her ratings and dynamics in McCain vote intentions is astonishingly exact. Her marginal impact in vote-intention estimation models dwarfs that for any Vice-Presidential we are aware of, certainly for her predecessors in 2000 and 2004. And the range traversed by her favorability ratings is truly impressive. But why? We are unaware of any theory that opens the door to serious impact from the bottom half of the ticket.
The paper is here.