Football games, shark attacks, and why voters may not be so incompetent after all

By Anthony Fowler

The wild tone and tenor of the presidential campaign has led many commentators to bemoan the state of our democracy. Often, these commentators have turned to political scientists to understand how voters behave, and some of the evidence appears quite troubling.

For instance, in a 2010 article in the Proceedings of the National Academy of Sciences, Andrew Healy, Neil Malhotra, and Cecilia Hyunjung Mo report that college football results influence support for incumbents in the home county of the team. And in a 2016 book, Christopher Achen and Larry Bartels report shark attacks hurt Woodrow Wilson’s 1916 reelection bid in New Jersey beach communities. Drawing upon these results, a recent Vox article suggests that democratic elections fail simply because voters are incompetent. If voters can allow arbitrary and irrelevant factors like football games and shark attacks to influence their choices, it’s no wonder elections feel more like beauty contests than interviews for the most important jobs in the country.

But before we express dismay toward the voters and the political process, we should assess the validity of these empirical claims. Is it really true that voters are so fickle, irrational, and incompetent that they allow football games and shark attacks to influence their vote for president?

Let’s start with football. Healy, Malhotra, and Mo report that college football victories benefit incumbent presidential, gubernatorial, and congressional candidates. Specifically, the outcome of the game 10 days before the election purportedly influences the incumbent party’s vote share by 1–2 percentage points in the home county of the team. Because this result is so surprising and because its implications are so dire, Pablo Montagnes and I reassessed its validity in a 2015 paper also published in the Proceedings of the National Academy of Sciences. We wanted to know if there was a genuine effect of college football games on elections. We couldn’t re-run the past fifty years of college football games and elections, so we tested additional independent hypotheses that should hold if the purported effect is genuine. None of these tests were consistent with a real effect. The estimated effect of college football games is actually smaller in counties where people care more about college football, no greater when the incumbent runs for reelection (as opposed to retiring), and no greater in the home county of the team relative to the rest of the state. And importantly, we replicated these tests for NFL games, which are about 10 times as popular as college football games, but we found no effect.

What about the sharks? Achen and Bartels report that fatal shark attacks in New Jersey in the summer of 1916 hurt Woodrow Wilson’s reelection bid in the beach communities. As with the football study, the shark result is supposed to demonstrate that voters are incompetent and that democratic elections do not work as hoped. Andy Hall and I reassessed the sharks finding in a recent working paper. First, we collected data on every fatal shark attack in the U.S. and every presidential election between 1872 and 2012, and we find no systematic evidence of an effect. Next, we ask what explains the New Jersey results from 1916. Achen and Bartels conducted two different analyses — one of New Jersey counties and one of towns within Ocean County, N.J. For their county level analysis, they made a number of seemingly arbitrary choices such as controlling for “machine” counties and deciding which counties to classify as affected. We show that virtually all of those choices are consequential for their results. Alternative choices produce results that are substantively smaller and statistically weaker. For their town-level analysis, we show that they did not appropriately account for town boundary changes and they mistakenly dropped one town whose boundary did not change. When we correct these errors, the estimated effect in the beach towns of Ocean County shrinks significantly, and when we replicate their tests for the other beach counties of New Jersey, we detect no effect. Lastly, we demonstrate that the statistical procedure used in the original shark study is prone to false positives.

Why did some of the most respected political scientists produce results that now appear to be false? One answer has to do with the incentives. Because of the attention generated by surprising results, researchers have a strong incentive to find them, and journals and presses have a strong incentive to publish them. The more the research community looks for unlikely effects, the more likely we are to discover them, regardless of whether they are genuine or not. And all of this can happen with perfectly honest and able researchers.

So what does all of this have to do with the wild 2016 campaign? Research suggests that voters often appear reasonable and competent. They generally respond to the economy and government performance (for example, here). They seem to care about policy and punish extremists (see here, here, here, and here). And they seem to select on quality, such that candidates who have won more elections perform better in office (for example, here).

That’s not to say that they’re perfect. It’s hard to figure out whether an incumbent is doing a good job, because there are so many factors outside his or her control. But voters might be doing a reasonable job with what little information they have. Furthermore, just because voters might hold controversial views or support a controversial candidate, that doesn’t mean that democracy isn’t working as intended. I suspect Trump is primarily getting support from people who genuinely want fewer immigrants and lower taxes for the rich, and voting based on policy goals — even objectionable ones — is neither irrational nor incompetent.

Regardless of what you think of the current campaign season, there is no evidence that irrelevant events like football games and shark attacks influence elections, and voters are probably more competent than one would infer from the headline-grabbing studies.

Anthony Fowler is an Assistant Professor in the Harris School of Public Policy Studies at the University of Chicago. He conducts research on elections and political representation, and his work has recently appeared in the American Journal of Political Science, Journal of Politics, Quarterly Journal of Political Science, and Proceedings of the National Academy of Sciences.