Two years ago – during the coverage of the 2012 presidential race – the relationship between Nate Silver’s Five Thirty Eight (then based at The Times) and the paper’s political team seemed tense. And an observant reader could sometimes pick that up in the journalism.
The data-based Five Thirty Eight had one strong narrative based on Mr. Silver’s algorithm: President Obama very likely would win re-election handily. Political reporters were presenting the race, based on traditional reporting and analysis, as either side’s to win. And Mr. Silver’s efforts were generally not integrated into the whole of The Times’s coverage.
This year’s midterm elections were a different story. The replacement for Five Thirty Eight (Mr. Silver left The Times for ESPN last year) is The Upshot, headed by a Times insider, David Leonhardt, who had been the Washington bureau chief. And Times coverage leading up to the election not only integrated what The Upshot was predicting but reflected what it was predicting: a big Republican win in the Senate.
On Election Night, The Upshot’s elegant and effective Senate election model was on full display, and very much a part of the whole.
On Wednesday, I caught up with Mr. Leonhardt, asking him how it went and what might be learned from the big night.
“This is the biggest news event The Upshot has covered in its existence,” he said. “We’ve been building up to this.” He said he was pleased with how it went, had some thoughts about how to improve it and gave particular credit to The Upshot’s Amanda Cox, Josh Katz and Nate Cohn.
The model’s transparency to readers (it published the code behind its formula and allowed readers to play with how the various factors were weighted) was one of its most innovative and best qualities.
To its credit, The Upshot, during the campaign, also presented what other models were showing – including Five Thirty Eight’s.
“We don’t think we’re the smartest kid in the room,” Mr. Leonhardt said. At one point during the campaign, he noted, The Times’s home page carried the logos of Mr. Silver’s outfit, as well as those of The Washington Post, Huffington Post and others. The Times wouldn’t always have done something like that.
He was also pleased with the use of the statistical analysis of voter files – the public record of registered voters and their choices – and with the Election Night coverage, which allowed readers to follow the vote count more closely.
“As a reader, I’ve long been frustrated with having no intelligent way to follow the vote count,” he said. The Upshot’s tracker (which adjusts based on where votes are in and where they are not) allowed readers at home to “move a lot closer to what the experts know, in real time.”
As for lessons learned, Mr. Leonhardt said he wants to find a way to get the tone right – “it’s important that we not oversell the clairvoyance of this approach.”
“The message should not be ‘we know what’s going to happen,’” he said. “We want to say, ‘this is better but not perfect — there are still surprises out there.’ Maybe that’s easier the second time around.”
And just how clairvoyant was The Upshot? Here are some reviews and comments, the last one of which (apparently referring to campaign models, not Election Night coverage) speaks to Mr. Leonhardt’s point: