Jonathan Chait graduated from the University of Michigan in 1994. If you follow him on Twitter or read his columns at The New Republic, you know that, in addition to his left-leaning politics, he’s also an avid college football fan. So why would this die-hard Michigan Wolverine, a man who once called Ohio State fans “insane hooligans,” root for his arch-rival? Well, as the old adage goes: politics makes strange bedfellows. A 2010 paper by political economists Andrew Healy, Neil Malhotra, and Cecilia Mo shows us that local college football games won by the home team prior to the election increases an incumbent candidate’s share of the two party vote. Simply put, if OSU continues its winning streak this weekend, numerous Ohio voters will head to the polls in a better overall mood. Healy, Malhotra, and Mo demonstrate this effect in presidential contests as well as Senate and gubernatorial elections. As the authors note: “These findings underscore the subtle power of irrelevant events in shaping important real-world decisions…”
What I wanted to do here is examine the magnitude of this effect with respect to Ohio State’s football team and Ohio electoral results. Close political observers know that Ohio represents the key swing state this election, so it's preeminent football program deserves special attention (sorry Bengals and Browns fans). In their wider examination, Healy, Malhotra, and Mo estimated the effect of on-field success and an incumbent's share of the two party vote to be between 1.6 to 3.0 points (depending on a handful of factors). This is within the bounds of Obama’s current lead over Romney (for example, Nate Silver has Obama at +2.7 in Ohio), making the size of this effect potentially meaningful.
So here’s what I did. I coded the two party vote in Ohio from 1976 to 2008 as well as the one year change in unemployment (the latter coming from the Bureau of Labor Statistics). I then recorded the Buckeyes' football record prior to the November presidential election. I ran three regression models using (1) OSU’s overall winning percentage; (2) OSU’s Big 10 winning percentage; and (3) whether OSU had a perfect season going into the election.
The results are in the charts to the left. The variable controlling for change in unemployment in statistically significant and correctly signed. This factor alone accounts for 75.1% of the variation in an incumbent president’s vote share, consistent with much prior research. And while none of the football variables are statistically significant, OSU’s Big 10 winning percentage and the perfect season indicator are headed in that direction (caveat: with only 9 observations). The addition of these two variable improve the model’s fit by 1% to 1.5% (respectively). But because the Healy, Malhotra and Mo paper uncovers a statistically meaningful effect with larger dataset and better specified models, here I’m most interested in the point estimates. The results suggest that a perfect season by Ohio State’s football team heading into the presidential election (which they currently have) is estimated in the third model to be worth an additional 1.3% of the two party vote. By comparison, a perfect season in the Big 10 (though not necessarily a perfect season overall) is worth an additional 2.0%. The scatter plots to the left contain those estimates with the 2012 prediction added. These two models are predicting that Obama receives about 56% of the two party vote (which translates into an overall popular vote of about 54%). This is certainly higher than what recent polling suggests will occur. And while this analysis is certainly crude, it is at least suggestive that OSU’s success on the football field this year may yield a marginal improvement in Obama’s electoral prospects. But just to be clear, the change in the unemployment rate is overwhelmingly driving factor behind the model's estimates.
Extra point: If Obama loses Florida, it’s probably because the Gators lost to Georgia last weekend. Or maybe not…
Editorial note: see Slate.com for a similar article.