Insights Into A Data Driven Election
With the 2008 Presidential election now almost a week behind us, the media is filled with backwards looking punditry on what lessons this campaign will inform history with.
But of all the unique aspects of this campaign, one thing that stood out was the use of data and how that influenced strategy, especially with the Obama campaign.
His unique electoral strategy of looking outside the typical swing states into areas where Republicans have always been strong (Colorado, North Carolina, Virginia) was driven by statistical analyses that showed how changing demographics in these typically Republican areas provided opportunities for a Democrat willing to take advantage of them.
I tend not to be very politically minded, but one site that fascinated me throughout the election season was Nate Silver’s FiveThirtyEight.com blog (538 being the number of electors in the Electoral College).
Throughout the campaign, he aggregated all the available polls and then analyzed them using regression analyses to find out what their outlier tendencies tended to be.
He then weighted the polls and re-simulated the election 10,000 times per update in order to, in his words, “provide a probabilistic assessment of electoral outcomes based on a historical analysis of polling data since 1952″.
And his accuracy throughout the election process was remarkable. According to this New York Times article entitled “This Math Whiz Called It For Obama Months Ago”, in the primary election versus Hilary Clinton, Silver “projected Senator Obama would win 833 Super Tuesday delegates, which was within about a dozen of the actual vote estimates”.
Additionally, when the returns came in on election night, it was found that “Mr. Silver had predicted the popular vote within one percentage point, predicted 49 of 50 states’ results correctly, and predicted all of the resolved Senate races correctly”.
What will be interesting to see is how this new approach to the analysis of polling data will have an effect on future elections. What is certain is that the data driven approach to election strategy is probably here to stay.