A few years ago, on a visit to the Topkap? Museum in Istanbul, I came across an amazing display of tunics that had been among the most tightly guarded treasures of the Ottoman emperors. They were not silk brocade, nor encrusted with jewels. The tunics were very plain, of white linen, but they had been inscribed with strange combinations of numbers and symbols. The Ottomans believed that they had talismanic power and that they could protect the sultan from a would-be assassin, turning back the blade of a knife or rendering harmless some deadly poison. Reading the inscriptions, I learned about Fazlallah Astarbadi and the Hurufis, a group of Sufi mystics who beginning in the thirteenth century developed the abjad (????) system under which sacred texts were deconstructed letter by letter and the individual letters were given numerical properties. There was a powerful relationship between these numbers, the Hurufis thought, a relationship that governed our lives. Fazlallah Astarbadi was, of course, following in the footsteps of Pythagoras, and of the Kaballists.
These days I start every morning with a sense that unfathomable numbers are governing my life. I strain to make sense of them. But often I give up, surrendering to the whirlpool of contradictions. I am talking, of course, about polling data. Somehow I think I can’t face the day without hearing Chuck Todd of MSNBC read the latest numbers, or watching them stream on the chyrons at the bottom of the tube.
But reading and understanding these polls—that is something limited to the true brotherhood of initiates. Fazlallah is right. Is there hope for the rest of us?
In the middle of this election season, I discovered Nate Silver and his website, fivethirtyeight.com. Nate lives in Chicago and is a writer for a sports-media company. He is also a polling nerd (baseball and politics are data-driven businesses, he tells us—but that’s an understatement). The number 538 represents the total number of votes in the American electoral college. Nate offers a state-by-state projection of how the presidential race is playing out, based on the polls. Of course, there are a number of sites that do this, such as pollster.com and electoral-vote.com. But Nate does them one better by offering a continuous tutorial on polling. What do the vexatious differences in those polls mean? What are the assumptions and models that produce those differences? Reading Nate through the fall, I feel that I have actually come to understand a bit about polling (at least, enough to realize that I absolutely must turn to a Nate Silver, Mark Blumenthal or Chuck Todd to really understand them). The process of learning is certainly a reassuring process, however.
Today, for instance, Mason-Dixon and Fox issued some polling numbers that suggested rather strong improvement for McCain. They were at odds with other polls, which generally showed a stable lead for Obama. How does one approach with these contradictions? Here’s a snippet from Nate’s explanation:
Mason-Dixon has also had a Republican “lean” this cycle of perhaps 2–3 points. They are quite frequently the most favorable number for John McCain in any given state. That doesn’t mean that they are “biased”, and it doesn’t mean that they are wrong–there are many different (and legitimate!) ways to think about this election. But it does mean that their polls need to be interpreted in that context.
Nate Silver is the pollster in my book. Election night, he’ll be setting up shop with Dan Rather on HDNet TV. That will be the place to stop in for number-crunching.