Editor-in-chief of FiveThirtyEight Nate Silver—widely considered one of the most reliable statisticians and election forecasters—said Wednesday that he’s “never seen otherwise-smart people in so much denial” about the chances of Trump winning the presidency. Silver followed the statement by defending his data showing Trump jumping from a 3% chance of winning just one month ago to a 48% chance now.
Early Wednesday PredictWise analyst David Rothschild ripped FiveThirtyEight for the massive swing in electoral odds, using its “roller-coaster” results as an opportunity to push his own firm:
“Something wrong with @FiveThirtyEight forecast. If you enjoy roller-coasters, please jump on, otherwise, followhttps://predictwise.com/politics/2016-president-winner …” tweeted Rothschild.
Silver hit back hard, accusing Rothschild of being—like so many analysts and political insiders over the last year—in “denial”:
“Never seen otherwise-smart people in so much denial about something as they are about Trump’s chances. Same mistake as primaries, Brexit,” he wrote.
Silver then issued a series of posts defending his group’s forecast of Clinton as the “narrow favorite,” which he notes is based entirely on polling data:
Our forecast has it close right now, with Clinton as the narrow favorite, because that’s what the polls show. It’s as simple as that.
The best (empirically) set of assumptions apart from the polls, that elections are driven by economic conditions, ALSO suggest a close race.
Can you invent an alternate set of assumptions wherein Clinton is a heavy favorite? Sure. But that’s despite the polls, not because of them.
If your premise is “Clinton’s only 1-2 points ahead right now, but she’s a fairly heavy favorite despite that…”, I’ll hear your case out…
…but people spend a lot of time denying & cherry-picking their way around the 1-2 points part, which suggests they’re not thinking clearly.
It’s not just 538. Simple, relatively assumption-free polling averages, such as RCP and @TPM, also show a very close race.
To put it another way, the burden of proof should be on people who DON’T show a competitive race, because that requires MORE assumptions.
Instead, there are a lot of *implicit* and *unexamined* assumptions that the race will shift back to Clinton. That’s dangerous.”
Rothschild responded to Silver’s defense by highlighting what he believed were the problems of Silver’s model: “@NateSilver538 the only *problematic* assumptions are (1) your house effects for st polling (2) your massive correlated error between st.”
Asked by Brendan Nyhan if other poll-based forecasts showed similar swings, Silver replied, “The @DrewLinzer / Daily Kos model is conceptually similar to ours. Has tracked 538 very closely (now 63% Clinton).” Silver later posted some of the findings of other groups as a comparison to his own company’s forecasts:
Below are PredictWise’s most current odds as of the publication of this article:
.png)
.png)
