Nate Silver's Wizardry and its Impact on the Media
Nate Silver predicted Obama was heading for an easy victory with a greater than 90% probability of winning all eight battleground states and 332 electoral college votes. This prediction, although shared by two leading academic pollsters (Drew Linzer, of Emory University and Sam Wang, of Princeton), was declared politically biased and blatantly mistaken by conservative experts who prophesied that the race was either too close to call, or that Romney would win.
The predictions of prominent conservative pundits included:
1. Peggy Noonan WSJ who wrote in her blog the day before the election that Romney would win because...”it just felt like it.”
2. MSNBC host Joe Scarborough who called Silver an ‘ideologue’ and pronounced the election too close to call.
3. Newt Gingrich who predicted 53% of the popular vote and at least 300 Electoral College votes for Romney.
4. Glenn Beck who predicted “321-217 victory for Romney in the Electoral College.”
5. George Will who saw a “321-217 Electoral College landslide, including a Romney victory in nearly every swing state...”
6. Larry Kudlow who confidently foresaw “...a landslide with Romney gaining 330 electoral votes.”
7. Dick Morris who looked forward to “a ‘dominant’ Romney victory with 325 electoral votes. His rationale: the media polls oversampled Democrats. So they're all wrong.”
8. Karl Rove, the ultimate operator, whose authoritative prediction was that “Romney would carry at least 279 electoral votes with a three-point popular vote margin” He compounded his error by his vigorous protests as Fox News called the victory for Obama:
As we now know, these impressionistic assessments proved to be worthless. In fact, the results matched Silver’s predictions, scarily perfectly. So, what has been the reaction of those who poured scorn on him? Are they issuing heartfelt apologies to Silver? Is there a widespread recognition that gut feeling is patently unreliable? Are we witnessing political consultants and talking heads taking courses in data science to better understand Silver’s methods? And are conservatives embracing data in the attempt to understand what really happened in this election? It would seem that the answer to all such questions is a resounding ‘no’! So what does this mean for these pundits and further, what does it mean for the media?
Perhaps not surprisingly, it appears to be hard for an old dog to learn new tricks. Confronted with their failure, the pundits have simply resorted to the same style of impressionistic ramblings that took the place of data science and analysis prior to the election. Much has been made, for example, of the significance of Hurricane Sandy that allegedly ‘broke Romney’s momentum’ and helped Obama ‘look Presidential’. “Hurricane Sandy saved Barack Obama’s presidency,” claimed former Mississippi governor Haley Barbour, “It broke the momentum that Romney had coming into the end of October.” Such statements, however, are made with the now familiar characteristic that they have no meaningful validation. Contrast Barbour’s unsubstantiated claim, with Nate Silver’s analysis of the Hurricane’s impact prior to the election, in which he marshals an impressive amount of data to argue that while there was some impact, it was small and overshadowed by larger longer run trends in Obama’s favor. See the difference? It seems that even when proven wrong, the pundits still find it very hard to break the habits of seemingly random guesswork.
Are there lessons to be learned from the data? Fortunately, it would seem that the answer is yes. The ‘Daily Beast’ expresses it succinctly. “In reality, the Republican Party didn’t lose the election because of Sandy, or Christie, or a mural. It lost because 71 percent of Latinos, 93 percent of black people, 73 percent of Asian Americans, and 55 percent of women voted against it. The party did not embrace policies that appeal to these demographic groups—and lost. And that’s the GOP’s fault.” Surely these pundits’ impressions, masquerading as analysis, are doing the Republican Party more harm than good.
Perhaps the broader question we should be asking is, ‘what is the role of journalism and media in the light of the evident intellectual bankruptcy of a host of familiar talking heads’? The answer would seem to depend on what we believe the purpose media to be. If it is to entertain; then pundits with opposing views provide low cost dramatic content. If however, it is to edify the public, then such fare is worthless. The age of the pundit prophet is over, and the age of the data scientist has arrived, it’s time we all recognize that and get onboard.
-Ben