Using data to make sense of nonsense
by Declan Murphy
Staff Data Analyst
Nate Silver is changing the game in politics—for better or for worse.
For those of you who are unfamiliar, Nate Silver is the head of FiveThirtyEight, a site that analyzes data and aggregates results to predict a variety of things. FiveThirtyEight covers topics ranging from sports wins and losses to economic forecasts, but what it’s most known for is its election and campaign coverage.
FiveThirtyEight first came to prominence in the ’08 presidential election, and while it has stayed relevant, this is perhaps the height of FiveThirtyEight’s importance. Of course, election years come with a predictable boost in traffic—but this election has been anything but predictable. The constant gaffes of Donald Trump, as well as the continual resurgence of investigations into Hillary Clinton’s emails, have created a chaotic and ever-changing public opinion. With its constantly updating polls and predictions, FiveThirtyEight has capitalized on this trend.
Strictly speaking, FiveThirtyEight does not create polls—it aggregates them. Silver takes in results from a number of pollsters and universities to update the ever changing picture. Silver also, somewhat controversially, adjusts these results to reflect any ‘house biases’. That is to say, Silver looks at the results of polls in context of any political leanings or standard data deviances to correct for any sources of potential error or bias. Then he weighs polls based on reliability, factors it all in to algorithms, and spits out a (more or less) holistic picture of what’s going on with the electorate.
It can be quite effective, if one agrees with his methods. Silver and his proponents argue that the sum is greater than the parts—that by incorporating the results of multiple polls, Silver creates the most thorough snapshot of what’s happening nationwide. (It’s also worth noting that, due to his number of sources, Silver’s presidential predictions also includes the odds of either candidate winning each state—something that national polls can’t efficiently recreate).
But detractors say that he undervalues the work of pollsters, is hasty to dismiss sources of ‘bias’ in polls, and profits from the work of others. Silver creates nothing new, in their view, but is all too quick to dismiss the findings of others.
Both sides have valid viewpoints. The court of public opinion, however, seems much less divided. Silver has become a bona fide celebrity. When, after the Republican convention, Silver’s “Now-cast” said Trump would win the election, it made headlines. (The Now-cast, it should be noted, does not reflect the long-term forecast, which still has Trump losing. It was merely the result of a post-convention surge in polls.) Silver also started a podcast, FiveThirtyEight Elections, that has sat in the iTunes Top 100 podcasts since it began in January. In fact, during the primaries, Silver’s podcast frequently made the top 10. Which is to say that Silver has become a sort of icon in the world of numbers, a hard-data guy with an easy to read take on the elections.
FiveThirtyEight is polling for the Internet era. With so many sources to drawn upon, and all this information at our fingertips, it doesn’t make sense to consult a single poll anymore. Indeed, the people who trot out single polls often point to the ones that support their own interpretation of events. It’s how Trump and his supporters spin things to make his chances seem greater. (For reference, as of this writing, FiveThirtyEight puts Hillary Clinton’s chances of winning the election at 71%.) But there’s a danger to it too. Devaluing the work of individual polls may someday rob Silver of his sources of raw data. If Silver wants to stay a leading voice in politics, he’ll need to starting recognizing that he can’t do it alone.