A Response to Seth Masket on the Impact of Forecasting on Elections

Political Scientist Seth Masket blogged in response to my musing last week questioning whether the rise of forecasting models might affect electoral outcomes. 

I wanted to respond to several of Masket's points. Overall, I agree with his characterization of the electorate. Most average voters don't pay any attention to the sorts of blogs/news outlets that report on forecasting models or the forecasting models themselves. 

I also agree that the sort of voters most likely to be deterred by forecasting models are more likely to vote in presidential elections. In fact, I did not intend for my original blog to raise the question solely in regard to the upcoming midterm elections. I saw it more as raising the question for now and the future. Forecasting is only going to increase as we become more interested in statistics and models in every aspect of society. 

I do differ from Masket on two points— first, I wonder whether forecasts, like polls, will receive any attention in local media (especially on local television news)? I see their outputs as the sort of simple, straightforward story that local news loves because it fits easily into a 30 second piece. For smaller news outlets, forecasts also offer the attraction of allowing for coverage of a race without necessitating costly reporting. 

I do not really worry that voters who actively pay attention to forecasting models might not vote. As Masket correctly notes, these are the sorts of "geeks" who are very likely to vote. What I worry about is people hearing about these results second or thirdhand, and/or through shorthand coverage in local media. "An updated forecast shows that Senator So and So only has a 23% chance of re-election," is the sort of sexy, uncomplicated headline that I worry that people who are less informed might see. These voters don't have sufficient information to judge the data in the model, nor are they likely to know any of the nuance behind that headline. 

Additionally, while Masket is correct that lower attention voters typically don't vote in midterm elections, Democrats' entire strategy seems to center around getting some of these voters to the polls. They have invested millions in field campaigns (as Phillip Rucker detailed with regard to Alaska yesterday), and clearly want to change the traditional midterm electorate. Whether this gambit works is anyone's guess (though I wouldn't rule it out given Sasha' Issenberg's findings about the sophistication of the Democratic microtargeting efforts). 

Masket may be correct— especially given early voting options that didn't exist before the last few election cycles— the rise of modeling may have 0 impact on who turns out to vote. But I think we need to know more about how the information/forecasts from these models diffuse before knowing with any degree of certainly whether they will or not.