In the chapter about political predictions in Nate Silver’s 2012 book The Signal and the Noise, FiveThirtyEight’s founder explains how there are prognosticators who fantasize about making a “daring, audacious, outside-the-box prediction,” but that predictions based on consensus, multiple sources with different approaches, tend to be more accurate.
Consensus is a major theme in interactive features about how the Senate may turn out following November’s midterm elections. FiveThirtyEight, The New York Times, The Washington Post and The Huffington Post offer some of the more visually stunning 2014 election forecast features on the Web, but as a whole, they all seem to be all saying the same thing: Republicans will probably take control of the Senate. (Though HuffPo and the Post, as of Thursday, moved the election close to a coin flip.)
This makes sense to a certain extent.
“All these media organizations are using pretty much the same information so it is no surprise that their predictions are similar,” said Andrew Gelman, a statistics professor at Columbia University, in an email. The outlets are all following the same poll results and the same news cycle. If the organizations all are following the right practices to build forecasts based on probability, they probably will come to the same result.
But there are some key differences in forecasting features, mostly in presentation and explanation. Through some nifty interactive features, both HuffPo and the Times get LAURELS for conveying that their predictions are built on laws of probability and that any number of scenarios can happen. On HuffPo’s page, visitors can play with a tool that allows them to calibrate outcomes in what are likely to be the nine most hotly contested races and see what bearing the individual races may have on the makeup of the Senate as a whole. The Times has a simple button inviting users to “spin” 36 Senate election roulette wheels. These interactives aren’t just fun; they get the point across that there are many factors at play in this election.
While not entirely divulging the proportions in their secret sauce, HuffPo, FiveThirtyEight, and the Times do a good job of explaining the interplay of the different polls used to make the election models. The Washington Post has an explanation for its model too, which includes qualitative factors such as “national landscape,” partisanship, and “key features of the race” alongside polling. The qualitative factors are all important features to consider, but given that the newspaper’s forecast is expressed in percentages and probabilities — rather than a simple Democrat versus Republican prediction — we do wish there were better explanation of how the factors are considered.
All of the sites examined in this story get DARTS for failing to chart the probabilities over time alongside context from the individual races and the nation as a whole. This may seem inconsistent with our criticism of The Washington Post for considering “key features of the race” while failing to explain how it considers those features. Context can’t be given a quantitative value equivalent to polling numbers, at least not without articulating a methodology, but it can be used to analyze the polling data. The Times does the best job of charting the swings in the overall Senate race since the beginning of the year, but it doesn’t explain why swings happen, including a recent one that nearly brought the Democrats even with Republicans. (The recent swing was caused by changes in Kansas’ Senate race.)
The main journalistic value in these features should not necessarily be to be 100 percent correct come Election Day, but rather to help explain how the nation arrived to its collective stance. By combining forecasting with context, news organizations can add value to these features, not only as the races grind to completion, but well after, when we have a chance to step back and examine the 2014 midterms as a historic moment.
Tanveer Ali is a Chicago-based journalist who is DNAinfo.com Chicago’s data reporter and social media producer. He has reported for the Chicago News Cooperative, WBEZ, and GOOD Magazine, among others. A former staff writer at the Detroit News, he received a master’s in journalism from the Medill School of Journalism.