The dramatic finish of the Iowa caucus — the confirmation of a late surge in support for Rick Santorum and the eight-vote margin separating him and Mitt Romney — brings up the question of what happens next. Here is one systematic finding:
What I’ve done is compare candidates’ performances in the Iowa caucus against the pre-caucus polls (the horizontal axis) to changes in media attention post-caucus vs. pre-caucus (the vertical axis). The polling data are from Nate Silver. The media data are from a new book by the political scientists David Redlawsk, Caroline Tolbert and Todd Donovan, “Why Iowa”? — which I previously blogged about at the Monkey Cage. They simply count the percentage of coverage for each candidate in The New York Times. The data include primaries from 1980 to 2008 and 76 different candidates.
The conventional wisdom is this: candidates who overperform then receive increased attention from the news media — presumably because they have exceeded the news media’s expectations, which we can approximate with the pre-caucus polls. This is exactly what the graph shows. For every three-point increase in Iowa caucus performance relative to polls, candidates can expect to gain an additional two percentage points of media attention. To be sure, the relationship is noisy and there are some outliers. Polls are not perfect proxies for expectations, after all. (Tom Harkin’s drop in media attention in 1992, and Bill Clinton’s increase, may stem more from Mr. Harkin’s “poor” performance as a native son.) Nevertheless, the relationship is substantively (and statistically) significant.
Why does this matter? Mr. Redlawsk and his colleagues demonstrate that not only do candidates who do relatively well in Iowa do better in New Hampshire — see also Nate’s analysis — but this shift in media attention may play the causal role. The media’s attention matters too, and their attention depends on how candidates perform versus expectations. Mr. Redlawsk and his colleagues then show that the results in New Hampshire shape the candidates’ overall share of votes in the primaries as a whole. So Iowa affects New Hampshire, and New Hampshire affects everything else — at least as best as we can tell from this limited set of presidential primaries and from data that show correlations rather than causation. They argue that this happens because some voters use the Iowa and New Hampshire votes to reassess the candidate’s viability, or chance of winning the nomination. Perceived viability then affects their preferences among the candidates.
So, as with previous candidates who have overperformed the polls, I would expect Mr. Santorum to get more media attention. Expect that to help him somewhat in New Hampshire — but probably not enough to win, of course. And whatever the results in New Hampshire, expect them to resonate in the weeks ahead.