How Analytics Shaped an Obama Win, Romney Defeat

The stories are slowly coming out of the election night landscape where an overly confident Governor and an equally confident incumbent President basked in the glow of a hard fought campaign.  As the election results trickled in one thing was becoming clearer by the hour, one candidate had correlated the demographics of the citizen vote particularly in key battleground states – and the other had not.

If 2008 were the election year that President Obama won with a solid push of social media, 2012 could be the year that the incumbent President had a competitive advantage with clear transparency of the voting population based on his use of advanced, cloud-based analytics.  Hosted largely by Amazon Web Services (AWS) as described in this New York Times piece by Steve Lohr, the AWS platform allowed for canvassing, massive phone poling, and real-time forecast projections which painted a very different view particularly of the ground game than what the GOP was predicting based on 2008 demographics.  A completely different ground attack resulting in a largely unpredictable election outcome that even as the key battleground state of Ohio was being called for the President, few Romney supporters would believe the numbers staring in front of them.

This time, the Obama campaign’s data center was mainly Amazon Web Services, the leading supplier of cloud services. The campaign’s engineers built about 200 different programs that ran on the Amazon service including Dashboard, the remote calling tool, the campaign Web site, donation processing and data analytics applications.

These data analytics applications gave the Obama team a clearer vision of the ground game which became painfully obvious in the key state of Ohio:

• The Obama team had a higher number of voters in their turnout model in the key metropolitan areas surrounding Cleveland, Cincinnati, and the college city of Columbus.  GOP models accounted for a lower overall turnout particularly with historically minority groups.  In short, the Romney camp had too small a denominator for their projections.
• GOP models also incorrectly accounted for the Libertarian candidate, Governor Gary Johnson, who took most of the 1.6% “other” vote in the Buckeye state.  Conventional thinking in the Romney camp was that these conservatives would “come home” on election night.  They didn’t. Advanced voter forecasting particularly in rural counties where the Governor won overwhelmingly would have showed that the GOP didn’t win “enough” to take the state.

By and large the final embarrassment came on Fox news when a disbelieving election night crew went to confront their own projections room for a final answer on Ohio.  The analysts – realizing they were the messengers about to be shot – sheepishly stood by their projections with “99.95% accuracy” in calling Ohio for the President.  At that point even the Fox News projection room was apparently working with more sophisticated analytics tools than even the Romney campaign had available.

Even national trends escaped the GOP which is cause for much needed reflection.  While the GOP took a larger percentage of the Caucasian vote in 2012 by 2% over 2008, the overall electorate had shifted by nearly 5% of population swing to minority groups, particularly Latinos.  While the Obama camp likely saw this trend emerging, the Romney camp was completely oblivious to the fact that by achieving their own campaign objectives it would cost them 2 million votes.

In the end one could argue that the campaign outcome in the final days would be changed much differently. However the Romney camp, bouyed by its own inaccurate portrayal of the ground game, believed that states like Pennsylvania and Michigan might be in play and diverted much needed funds for advertising and events into those markets.  Romney played the safe “incumbent” role in both Ohio and Virginia since their forecast models showed victories in those states – as well as Florida.  In each case the suburban and urban vote came out strong for the President.  Again, Team Romney was simply working with the wrong denominator.

Much can be said about the candidate profiles and the Republican party will spend months dissecting what went wrong and what should have been.  In the end though, their strategies and ground tactics were flawed due to an inadequate use of available data analytics and forecasting technology which cast a very unrealistic image of what was actually happening in the nation.