Polling guru Nate Silver built his career sorting through election polling data with remarkable accuracy. 2016 put that reputation to the test.
Silver’s forecasting site FiveThirtyEight, named for the number of votes in the electoral college, went into the election giving Hillary Clinton a 71.4 percent chance of winning. The number was lower than some competing polling organizations, but still quite favorable.
{mosads}While 71.4 percent sounds like a pretty good bet to most, statisticians see those odds as predicting a loss roughly three out of ten times.
“I felt like, in a weird way, there wasn’t that much at stake, because I thought we were going to be criticized either way,” Silver said.
As the results came in at the FiveThirtyEight headquarters in New York, Silver said the statisticians were less shocked than the pundits.
“It wasn’t that big a surprise. I think it’s a persistent myth about the 2016 election. This was well within the range of polling errors that we had in past elections. But the polling error this time was enough to flip the Electoral College,” he said.
Silver blames the broad sense that Clinton was a shoo-in more on the media’s statistical illiteracy than on polling mistakes. If the result has lowered people’s confidence in polling, Silver said that’s their problem.
“To me, the notion that we can’t trust the polls, and we should just trust our gut instinct, that’s bullshit,” he said.