Journalists don’t like being wrong, or even wavering in the vicinity of not right. And in the run-up to Donald Trump’s 2016 victory, many of us were very wrong. Carefully designed visualizations fritzed out in the face of uncertainty.
So this year, for these complex midterms, media companies have adopted the posture of humble supplicants to the American voter, waiting on a greater percentage of votes to be cast in a greater percentage of races before making the call on the one big, plausibly contested thing that matters: Who will control the House?
The New York Times—whose skittering needle in 2016 was so anxiety-making it became a viral Halloween costume two years later—tweaked its strategy after “many, many” internal discussions. As a result, the paper refused to unleash a visualization until 9:45 p.m. ET. By 10:05, the needle showed that Democrats had a 95 percent chance of winning the House. Just minutes before, the page had displayed a diffident message. “We do not yet feel confident enough in our estimates to publish a live forecast,” it read. “If and when we do, we will publish it here.”
FiveThirtyEight, which prides itself on its data-centric view of politics, saw its own model moving too deeply rightward after early election results, and reset it to a more conservative (prediction-wise, not politically) approach. “After it was wobbling back and forth too aggressively early in the evening, we have the model tuned to a conservative setting, where it’s mostly just waiting for called races,” the site’s editor in chief, Nate Silver, wrote.