At FiveThirtyEight this week, statistics guru Nate Silver wrote a major apologia about his 2016 predictions called “How I acted like a pundit and screwed up on Donald Trump.” Its ostensible purpose was to explain a string of apparent misdiagnoses from a statistician famed for getting things right.
Inadvertently, though, it explained a lot about what we campaign journalists in general have done wrong to pave the way for the seeming outlier of a Trump nomination.
Silver’s voice is a big one in our business. After calling the presidential election for Barack Obama eight months early in 2008 and predicting 49 out of 50 state results, he re-wired the minds of a generation of campaign reporters. An ideological descendant of baseball statistician Bill James (whose Baseball Abstract was an annual purchase in my home growing up), Silver retrained political analysts to think in Moneyball terms, tuning out statistical noise and focusing on the actual path to electoral victory. The naming of his new site, FiveThirtyEight, was symbolic of this new emphasis on what mattered in presidential races, i.e. electoral votes.
A candidate who scored seemingly encouraging results in national polls while performing poorly on the state-by-state electoral map was the equivalent of the baseball player who hit for high average but didn’t draw walks or hit for power.
With the analyses of Silver and his crew, reporters now had a much better grasp of who was actually winning races, especially primary/nomination races, which relied on arcane delegate rules that Silver made it his business to understand.
On the campaign trail, the success of FiveThirtyEight inspired heated in-plane debates. More than once I had to suppress a laugh listening to a reporter grumble that campaigns would now be less about “the issues” thanks to Silver and his lot, as though campaign coverage hadn’t been 99.9% percent a horse race already.
Then 2016 happened. The normally cautious Silver howled from the rooftops that Trump was a temporary phenomenon. As late as November 23 of last year, he wrote a piece called, “Dear Media, Stop Freaking Out About Donald Trump’s Polls” that among other things noted that Trump’s numbers at the time — about 25 percent of the 25 percent of Americans who identified as Republican — roughly matched the number of Americans who believe the moon landing was faked.
Silver’s site repeatedly put hard numbers on Trump’s chances of victory. The results weren’t pretty, as he notes:
“In order of appearance — I may be missing a couple of instances — we put them at 2 percent (in August), 5 percent (in September), 6 percent (in November), around 7 percent (in early December), and 12 percent to 13 percent (in early January).”
To diagnose what went wrong, Silver among other things went back to see what the numbers said about other “Trump-like candidates.” He defines those as “candidates who led national polls at some point in the year before the Iowa caucuses, but who lacked broad support from ‘party elites‘ (such as measured by their number of endorsements, for example).”