previous article
The duel between Barack Obama and Mitt Romney in the last presidential election ended with a very clear result, a victory for Obama and a crashing defeat for Romney. The participants of my webinar regarding the topic know that this was not only due to the candidates` rhetorical skills, the climate in the American society and the political programs of the rivals, but that there was another big difference.
Predictive Analytics as Success Factor in the Election Campaign
This difference was Obama's very massive, very expensive, very professional and very successful use of a whole bundle of predictive analytics methods in the election campaign. These methods resulted in the fact that Obama's campaign supporters knew at any time on which doors to knock and which topic on Obama's political agenda to focus on in a conversation with the residents. They were able to exactly calculate how to persuade certain Obama supporters to vote and to pull waverers onto Obama's side. At the same time, Romney's supporters without comparable information wasted their time in streets where, using predictive methods, they would have known that there mainly are die-hard Democrats, which none of the campaign supporters would have been able to pull on Romney's side.
Statistical models also enabled Obama's team to precisely estimate in which of the "Swing States" they had already won the race anyway. Thus, they were able to concentrate their resources on the remaining states. On the election day it turned out that these models were so precise that they forecast the right result for every single state with a deviation of no more than one percentage point. While Obama's data science team celebrated the election day as "model validation day", Romney's men had nothing comparable to deploy.
Unique Data Basis
One of the strengths of the Obama campaign was the data basis on which they built up their various models. Unlike in most European countries, there is easily accessible and comprehensive data on who gave his/her vote at which election in the USA. This so-called voter file, an American specialty, was supplemented by Obama's team in a way which could hardly be copied in Europe: in the USA, data on the voters´ political preferences can be gained from the registration as supporter of a particular party. This registration is important and relatively popular as it entitles to participate in the pre-elections in which the parties select their candidates. In addition, Obama's team used bought-in marketing data which is also available in extreme variety and level of detail in the USA. In addition, comprehensive own telephone interviews were conducted, which only a financially strong organization such as the Obama campaign can afford.
This unique data basis enabled extremely helpful and valuable forecasts. The modelling methods used for this purpose were innovative in particular in forecasting the election results on state level. For managing the election supporters' assignments, models inspired by usual direct marketing standards were sufficient to manage the Obama campaign with unmatched accuracy.
What the Republicans Want to Improve in the Next Election Campaign
The Republicans drew the conclusions from this defeat. Amelia Showalter's presentation at the Predictive Analytics World 2015 in Berlin indicated that all Republican candidates for the next presidency formed a high caliber data science team in order not to repeat last time's data science disaster. All candidates. Except for one: Donald Trump.
