This week, ESPN and other sports news outlets carried an AP story which contained some actual political science. The original AP story summarized a study by Healy, Malhotra and Mo which found a link between the success of college football teams in the week leading up to Election Day and an incumbent’s vote share. From the study: “we find that a victory in the ten days before the election causes the incumbent party in Senate, gubernatorial, and presidential elections to gain an average of 1.6 percentage points of the vote in the county.” Obviously ESPN picked up this story because it presents a nice stylized fact about the effect of college football on politics. But this study has important implications for our understanding political behavior because it builds on previous work that suggests retrospective voting is conditioned by information that has little to do with an incumbent’s governance. In other words, voters appear to behave irrationally. Unlike prior work, this study analyzes events “entirely separate from public policy.” The authors posit that the causal mechanism is well-being—victories on the gridiron enhance a voter’s overall sense of well-being which increases support for the status–quo. Overall, the study is quite good. And kudos to mainstream media outlets for disseminating quantitative political science. If you read the actual paper, as I have, you will see that it is conceptually parsimonious and methodologically sophisticated. I have no doubt about the authors' central claim. That said, I had some thoughts while reading the piece.
First, as college football fan, I take issue with one aspect of the authors’ methodology. They restrict the sample to BCS conference teams, plus Notre Dame. They include the Fighting Irish because the university is “an independent school with a rich football tradition.” Clearly, they have erred. If I were reviewing this paper at a journal, I would insist that Notre Dame be excluded from the sample after 1996, the year Bob Davie began coaching.
I was also wondering about the supposed disconnect between government performance and success on the football field. Again, the Healy, Malhotra and Mo study pushes the boundaries because it considers a stimulus entirely separate from public policy concerns. I think this is a safe claim, but not perfect. Public colleges and universities are, after all, public. The question is: To what extent does the American citizenry perceive colleges and universities as government institutions? Though maybe a stretch, it could be that the behavioral outcome observed by the authors is driven, at least in part, because people are less upset with government education spending when their favorite college football team wins. Ok, that’s a big stretch. Still: What about private colleges? It would be interesting to analyze their results separately for public and private universities. Of course, I’m not sure most people grasp the distinction between the two types of institutions. Still, if the effect they uncover is diminished for private universities then some small component of their finding is driven by respondents linking college football with government action.
As someone interested in US political parties, I was interested in differences across majority and minority incumbents. Retrospective voting posits that individuals evaluate the government's past performance when making political calculations. But that assessment is uneven. When government is performing poorly, it’s the majority who typically receives the blame. As with my previous comment, if there are differences across parties, with the majority party receiving the most gain from college football victories, then we can get some leverage over the question about whether voters are making a direct connection between public policy and college football or a more superficial connection (with the latter showing up in the empirical results as a failure to reject the null hypothesis). Of course, the question of whether there is an actual link between government effectiveness and football success is irrelevant, so long as enough individuals believe one exists to produce a statistically meaningful difference.
I’m also curious about the effect of football sanctions. Sadly, the N for this question is sufficient for statistical analysis (USC, Alabama, etc.).
Finally, the authors might gain some leverage on the exact causal mechanism by thinking about Zaller’s work. In my mind, part of the causal process may have to do with humans’ limited information processing capabilities. The “well being” experienced by a college football victory is going to be short lived, I think (either 3 days or 10 days, Saturday to Tuesday, according to their methodology). Instead, the victory may “de-prime” political information (see Zaller’s A3 axiom). Thus, when voters go to the ballot box after a big win it may be that they are relying on less meaningful cues, say name recognition, which advantages incumbents. This inability to recall politically salient information is most pronounced if heavy tailgating was involved.