A Predictive-Descriptive Artificial Intelligence-based expert computer system predicts a Democratic landslide victory in the 2016 presidential race, regardless of the nominated candidate. The GOP would require a dramatic shift in the electorate to win the White House.
To win the 2016 U.S. presidential election, 270 electoral votes are required out of a total of 538 votes. The Artificial Intelligence (AI)-based, automated expert computer model predicts a potential Democrat win with 324 electoral votes, versus 214 for the Republicans. The AI-based computer model process the prediction in two phases. In phase one, the model predicts that 257 electoral votes can be confidently assigned to the Democratic candidate, while 201 electoral votes can be confidently assigned to the Republican candidate. In phase two of the prediction, the remaining 80 electoral votes in five swing states are assigned based on the underlying patterns and statistics learned by the computer and the probabilities captured by the model.
Without using the AI-based computer model, the current political alignment of the country predicts that 251 electoral votes can be confidently assigned to the Democratic candidate vs. 109 electoral votes to the Republican candidate. The remaining 181 electoral votes are assigned to a bigger pool of the so-called swing states. Alternatively, our AI-based computer system with human expert intervention and interaction predicts a Democrat win with 315 electoral votes, to 223 votes for the Republicans.
In July 2015, Moody’s Analytics predicted a Democratic win with 270 electoral votes, to 268 for the Republicans, regardless of who wins either party’s nomination. However, in their revised model in August, Moody’s suggested a small change in the forecast data that has swung the outcome from a statistical tie to a Democratic landslide win. Moody’s model, which has successfully predicted every election dating back to 1980 (including a perfect electoral vote prediction in the 2012 election) predicts that a Democratic candidate will win in 2016 with 326 electoral votes (against only 212 electoral votes for the Republicans). Moody’s will update its prediction every month running up to November 2016. The updated Moody’s Analytics presidential election forecast in October predicts the Democrats having a distinct advantage, with a main focus on five swing states; Florida, Ohio, Colorado, New Hampshire and Virginia.
While Moody’s Analytics election model predicts a Democratic electoral landslide in the 2016 presidential vote, a computer model built by Reuters predicts a Republican victory instead. The Reuters data model takes into account historical trends as an important factor, while largely ignoring the electoral system as a crucial variable. In addition to the Reuters computer system and Moody’s Analytics, the PredictWise project from Microsoft Research predicts a Democrat win in the 2016 presidential election with 57% of the electoral votes (306 electoral votes), to 43% (232 electoral votes) for the Republicans. In contrast to Reuters, both Moody’s and PredictiveWise predict a Democratic victory in 2016 and a quiet “rough tough hard Rodeo ride” for the Republican candidate.
We have simulated all the possible statistical scenarios – of the six potential Democratic nominees (Clinton, Sanders, Biden*, Webb, O’Malley, and Chafee) and the nine potential Republican nominees (Trump, Carson, Rubio, Bush, Fiorina, Cruz, Huckabee, Paul, and Kasich). As an average, our model predicts a Democrat win with 296 electoral votes, to 242 for the Republicans, with a potential Democrat landslide win, as high as 337 electoral votes for the Democrats and 201 electoral votes for the Republicans. In 2012, President Barack Obama won the 2012 presidential election with 332 electoral votes while Mitt Romney, his Republican opponent, garnered only 206 electoral votes.
According to Moody’s, the key swing states for 2016 include Colorado, Florida, Ohio, Virginia, Iowa, New Hampshire, Nevada, Pennsylvania, and Wisconsin. While Moody’s model predicts that three states account for the change in margin (with Ohio, Florida, and Colorado swinging from leaning Republican to leaning Democrat), the AI-based model predicts up to seven potential swing states: Colorado, Florida, Nevada, Ohio, Virginia, West Virginia and possibly Arkansas (low probability). In addition, Moody’s model estimates that three Republican candidates will influence the election results in Ohio and/or Florida, potentially making the outcome of these important states even more unpredictable. In contrast, the AI-based model predicts that five Republican candidates and two Democratic candidates will influence the election in five potential swing states.
Furthermore, the AI-based computer model predicts three major factors for a Democratic landslide victory, 1) the President Obama effect, 2) the President Clinton effect, 3) and the economy and the decline in unemployment rate. Other potential factors that are considered in the human expert intervention computer model are 1) the current and former governors in each swing state, 2) the most recent U.S. Congress election, 3) the unemployment rate trends, 4) and the relationship between a potential presidential nominee (as well as his/her views and background) and the swing state. All of these factors were included into the AI-based expert computer model using an AI-based expert rule system, using data such as the history of the presidential election and a few common sense rules extracted from basic general elections.
As far as unemployment rates, we used the unemployment rates of all U.S. counties to predict future unemployment rates for each county, based on the presidential nominee’s party affiliation and the most optimistic and pessimistic scenarios. The predictions are based on Machine Learning- and AI-based Predictive Analytics. We could further refine the model by including Governor, State Legislation, and Senate/House elections at the county and state levels.
Follow me on Twitter and add me on LinkedIn and Facebook for more posts about AI-based Predictive-Descriptive models and predictions. Visit my blog for further details on these results and other similar studies.