Why Horserace Coverage is Bad

Updated 10/6: Thanks to Seth Masket, I've  added a reference to a scholarly study on the 1980 election. 

Let me start by noting that I love the political junkies in the media. I think Chris Cillizza (The Fix), Chuck Todd, and others do a fantastic job. 

This also won't be a screed about how covering politics reduces coverage of policymaking, and public policy, and how it hurts our ability to address national ills (though I agree with this sentiment to some extent in off years). 

Rather, my concern is that our new fascination with forecasting and prediction may help shape electoral outcomes. Electoral forecasting used to be the purview of a few specialized practitioners and academics like Charlie Cook, Stu Rothenberg and Alan Abromowitz. 

Now, in the wake of the success of Nate Silver, there seem to be countless models predicting doom for the Democrats in November. 

As I watched my twitter feed chatter breathlessly about a surprising new Iowa Senate Poll on Saturday and what it meant, I started pondering the ramifications of this sort of chatter. 

Famously, in 1980, the networks infuriated Democrats when they called the presidential election before the polls closed in the West. Democrats lost many Congressional elections in the mountain west and western time zones, some by razor thin margins (Senate Foreign Relations Committee Chairman Frank Church lost by a mere 4,262 votes), and one theory was that people stayed home once they heard that the presidential election was decided. Scholarly evidence from Michael Delli Carpini, among others, supported this contention. 

Again, in 2000, Republicans were furious when networks initially called Florida for Al Gore before polls closed in the conservative Florida panhandle (which sits in the central time zone). 

My thought is that the new modeling craze could have a similar effect (this is a hypothesis that scholars could and should test). 

If you're a diehard partisan or an attentive voter, odds are that you are likely to vote regardless. But if you're one of the many Americans who cares less about and/or focuses less on politics, and you hear that your Senator has a 20% chance of reelection, it becomes a lot easier to justify prioritizing other things over voting. 

I worry that we've created a self perpetuating and deterministic process. Pollsters rely on likely voter screens in an attempt to be more accurate. Commentators ignore the importance of margins of error when discussing polls and don't explain how likely voter screens are determined. Models have great success predicting elections, and we take their predictions as the gospel and let them influence our behavior. 

By the time the average, detached voter hears about the results of polls and models, all nuance is lost. They might hear a 20 second headline on the local television news, or a friend might say did you hear that Senator X is going to lose?  

The end result is that we might end up with an electorate that consists more greatly of avid political followers and partisans. These segments of the electorate may be more polarized than  more casual voters. They also may overrepresent certain demographic groups and certain perspectives. 

One might scoff at this hypothesis. After all, if you care about the country why would you let a model influence whether or how you'd vote? And yet, we have study evidence that "rain significantly reduces voter participation by a rate of just less than 1% per inch, while an inch of snowfall decreases turnout by almost .5%." 

If the weather affects how willing people are to vote, is it really far-fetched to think that hearing that a voter's preferred candidate has only a small chance to win might reduce his/her likelihood to vote? 

One thing that many of us who love and study politics forget is that many Americans don't care for politics, find politicians tiresome and trite, and are very busy. Thus, they focus on politics only at key moments and when the stakes seem to be highest. If they tune in and see that the election won't be competitive, they might well tune back out. 

The skeptic might say, okay, but surely this coverage won't keep too many people away from the polls. My reply is that it could determine a razor thin margin of victory (and we might be headed towards some very close margins in Senate races). 

In an ideal world, the press would focus on substantive differences in campaigns, and focus less on trying to predict the future. As Hans Noel recently pointed out, we'll know who will control the Senate in just over a month. 

But the press exists in a commercial environment, and forecasting fits nicely with the statistical modeling craze that is proving to be quite popular with consumers. Likely, that means we're stuck with forecasting going forward. But commentators should be careful about overstating the meaning of polls and models. Let the voters decide the elections.