The Art of Losing: A FiveThirtyEight Story, Part II
2017-2022: Silver's last crusade, and polling for patriots.
As the dust from the 2016 election settled, FiveThirtyEight entered a state of denial that it has never really left since. Silver, remembering how often he came to the defense of Trump’s chances, and clearly bitter that people were accusing him of blowing the election when his model only dealt in probabilities, refused to admit fault. The people really to blame were the pollsters, who gave him bad information, or other forecasters, who were far too confident in Clinton, or his critics, for being too stupid to understand how probabilities work. He was trying to have his cake and eat it too: letting people interpret his probabilities as hard-and-fast calls when it benefitted him, but insisting that they be seen as unfalsifiable when it did not.
These self-serving post-hoc explanations obviously did little to convince the site’s doubters. So, whether they liked it or not, Silver and his staff were forced to enter the Trump era bereft of the illustrious reputation it had enjoyed before 2016. They would find their footing early on by going back to their roots of data visualization, putting together pages representing things like Trump’s approval rating and congressional voting records. These projects would all be done quite well. Throughout the Trump presidency, they would be regarded by experts and amateurs alike as the gold standards within their respective categories. But as for their actual coverage, FiveThirtyEight would find itself in a totally different world.
The site had always had a predominantly liberal audience, something that Silver, a self-styled libertarian “in between Gary Johnson and Mitt Romney,” was very well aware of. Throughout the Obama administration, he had acted as a sort of apolitical straight man to his relentlessly partisan readers, providing them with the satisfaction of hearing someone who made pains to show how he only cared about the numbers say things they wanted to hear anyways. After 2016, they would stick to the same style. The site would also avoid any editorializing of Trump’s administration, simply reporting on the state of it and the electoral implications of its actions, without providing any actual opinions.
Maybe this was noble. But it also could not have been more out of place within the context of the massive popular mobilization on the left under Trump. Liberals were outraged, on the warpath to defeat a man they utterly hated at any cost. They saw the country falling apart before their eyes and they wanted to have their intense emotions validated. The last thing they were interested in were lengthy analyses about which party “was losing the retirement primary.” If FiveThirtyEight’s style ever had a moment, the Trump era was not it.
And in more bad news for the site, the upcoming election in 2018 was not shaping up to be the kind of race where they could win their reputation back with a flashy, against-the-grain call. Democrats, with massive leads from the polls and history on their side, were very strongly positioned to win back the House. But Republicans, benefitting from a comically favorable map, were set to hold, or even expand, their Senate majority. Everyone knew exactly what would happen. There was no glory to be gained.
This was highly unfortunate for the site, as 2018 would also be the year that they would have perhaps their best forecast ever. Their House model would nail the actual results down to almost the exact seat. But even this would be somewhat tarnished by the site’s election night tracker going haywire and incorrectly giving the GOP overwhelming odds to keep their majority. Such was their luck under Trump. Even when they got things perfectly right, they still looked wrong.
After 2018 was, of course, 2020: Donald Trump’s re-election year. The race, standing to have monumentally high interest, must have looked to be a perfect staging ground for FiveThirtyEight to find its way back into the national consciousness. Like they did before 2016, they started off a year early with constant subjective evaluations of the upcoming open primary. And just like they did before in 2016, they found themselves utterly incapable of understanding what it was voters wanted. Much of their initial coverage would be odes to the aptitude of the likes of Kamala Harris and Beto O’Rourke, neither of whom would even make it to Iowa. Once again, you’d probably have done a better job at predicting the primary by going by the exact opposite of what FiveThirtyEight was saying. They should consider themselves very lucky that their coverage of the 2020 Democratic primary received so much less attention than their 2016 Republican primary coverage. While it wasn’t exactly quite as bad, it was pretty close.
The site would continue to labor its way through the year until it was time to officially release their actual model for the primary. This model proved to be serviceable, if not very prescient. Its main prediction before the entire party consolidated around Biden was that there would be a contested convention, resulting in much of their coverage for the primary being about something that would never even come close to happening. It wasn’t a great start.
But any worries that this shaky primary performance could jeopardize the site’s chances to make a reputational comeback would soon be rendered irrelevant. Right as the 2020 election started to begin, the world began falling apart.
Keep reading with a 7-day free trial
Subscribe to Ettingermentum Newsletter to keep reading this post and get 7 days of free access to the full post archives.