I’m still scarred from 2016. With an abiding faith in polls and data visualization, I trusted, like so many other people, that Hillary Clinton was a shoo-in for the presidency.
But even the best minds got it wrong. The polling was off by greater margins than ever before. And for all the maps, meters, and calculators that seemed so easy to read, there was an unintended consequence: The news outlets failed to depict the margin of error behind polling data—a margin of error that Hillary’s seemingly epic lead was actually well within. (You can read our full breakdown of what happened here.)
So I’ve abstained as much as possible from following the latest election infographics this year. However, as we approach election night November 3—knowing full well that because of mail-in voting, we might not actually know our next president by the end of the night—I feel the itch coming back. I need to follow the election results. You need to follow the election results. Who should we be following?
It’s a question that I posed to three prominent data designers: Which publications are they referencing to stay up-to-date, and who will they be obsessively refreshing on election night? Here are their answers.
In 2016, many people followed FiveThirtyEight religiously, as the site’s founder, Nate Silver, made his name calling the 2008 election for Obama months early. Eric Rodenbeck, founder of the cartography-focused data viz firm Stamen, was among the site’s fans who had their hearts broken by the actual election results.
“Despite a feeling of trauma and trepidation from the last time, FiveThirtyEight is still my go-to,” says Rodenbeck. “I like that the whole map updates when you pick a state winner. It uses your input to narrow things down to just the simulations that fit your inputs. There’s a video on how to use it that’s pretty informative.”
As Rodenbeck explains, FiveThirtyEight’s map allows you to analyze current projections, then tap on individual states to see what happens if Biden takes Florida or Trump takes Michigan again. And if you want a bit more insight into the map’s invisible logic, data viz expert Adam Pearce actually compared FiveThirtyEight’s model to that of The Economist. It’s a short, geeky read that might not leave you any more certain about which resource to follow this election, but it does work well to highlight how advanced statistics can boil down to reading tea leaves.
Is Rodenbeck any more skeptical about FiveThirtyEight this year than in 2016? Maybe not, but he is more aware of his own, natural biases in assuming that Biden’s current 88 in 100 chance of winning really means 100%.
“The big change for me is that I know that wild and wacky out-of-band things happen all the time,” says Rodenbeck. “A 10% chance of a Trump win? Previously I would have thought ‘Ah, that’ll never happen, don’t worry about it.’ But if you thought there was a 10% chance of there being a catastrophic earthquake today, you’d think twice about leaving your house.”
270 to Win and ‘The Economist’
Dino Citraro, the CEO and cofounder at the “do good with data” firm Periscopic, has difficulty choosing just one go-to. He casts another top vote for FiveThirtyEight, calling it “charming, accessible, and surprisingly deep.” He goes on: “I use it when I just want a quick glance at the projections. I love the idea [that it offers interactive] simulations versus just the polls.”
Citraro also likes 270 to Win, which he has followed for several election cycles. Aside from its main interactive maps, which are thematically similar to those of FiveThirtyEight, Citraro likes 270 to Win’s state-by-state breakdown of the latest polling, which eschews national maps for individual circle graphs that clearly show polling lead by state. “Finally, a good argument for pie charts,” he jokes.
Meanwhile, The Economist is Citraro’s favorite at-a-glance update, which presents election odds in clear language right up top before it rolls into any visualizations at all.
‘The New York Times’
For Giorgia Lupi, a partner at the design firm Pentagram who is known for bringing a human element to data, the best data visualization of this election is not about clarity or interactivity. It’s actually about presenting information in the proper context. “I’m a visualization expert, but fundamentally, I always think about how the general public will [react] based upon the context,” says Lupi. “It’s about how people relate to the numbers they see . . . [from] the polls to what’s going on in their life.”
As a result, she prefers treatments where data visualizations are integrated into deeper stories that spell out the significance of the visualization. “I’ve particularly enjoyed the interactive stories of The New York Times, where I think they’ve been really able to make the case for a bigger poll of data, but through stories people could read and relate.”
Lupi isn’t talking about the publication’s hated needle, a meter that wavers between candidates like a pressure gauge about to explode, that may or may not be back on election night. She’s specifically referring to some of the Times‘s excellent recent work, like a timeline of Trump’s finances, which walked readers through the numbers of his tax returns step-by-step, and a map of the two Americas financing the Trump and Biden campaigns.
These treatments aren’t necessarily go-tos for election night (yes, Lupi refreshes FiveThirtyEight for her quick updates too), but they do lay important groundwork for this moment in time, which can easily be lost in a race to count electoral votes. And that extra context can help fill the gaps that polling inevitably misses.
“My biggest point about what I feel could have been done better around the  elections is showing uncertainty, and that’s a big thing to me,” says Lupi. “There’s so much we don’t know, so many people who don’t answer a poll, so many people who might not change their mind but don’t go voting. . . . That’s a missing data point that I think needs to be treated like a data point.”