Timothy Shenk wrote for the Nation about what political scientists and policy journalists often miss about American politics:
As I write this sentence, Donald Trump has a 14.8 percent chance of becoming president. At least, such is the verdict of the “polls-only forecast” issued by Nate Silver and his team at FiveThirtyEight. Switching to the site’s “polls-plus forecast,” which accounts for factors like economic performance and electoral history, raises Trump’s odds to 18.3 percent. The New York Times offers a more reassuring estimate: According to its model, the Republican nominee has a 11 percent chance of moving into the White House next January.
These predictions would be more comforting if their creators hadn’t spent much of 2015 insisting that Trump would never make it this far. In September of last year, Silver estimated the likelihood of Trump’s winning the nomination at“about 5 percent.” That was up from the2 percent chance Silver gave him a month earlier, when he outlined the “six stages of doom” awaiting the billionaire’s campaign. FiveThirtyEight was just as skeptical of Bernie Sanders’s candidacy, writing after his spring 2015 announcement that the Vermont socialist had “pretty much no shot of winning.”
The data connoisseurs at FiveThirty¬Eight were far from the only figures who failed to see the support for Trump and Sanders coming, but the limits of their prognostications were particularly glaring because of their previous records. Silver had racked up an impressive set of accurate predictions since his 2008 debut as a political forecaster. He accurately called all 50 states in the 2012 presidential race, and while his performance was shakier in 2014, it was still imposing: He correctly projected 34 of 36 Senate elections and 31 of 36 gubernatorial races.