For the past 12 weeks, UI Professor SHELDON JACOBSON provided The News-Gazette with an updated snapshot of the race for president. He filed one final report Wednesday:
Donald J. Trump has been named the 45th president of the United States. In baffling public pundits and prognosticators, from his unexpected journey through the primaries to his defeating Hillary Clinton, Trump defied the odds in garnering the necessary number of Electoral College votes to win the White House.
Although one can easily criticize the polling data used by Election Analytics and other sites to make predictions, two observations can be made when reviewing what occurred.
As Election Day approached, voters gravitated away from the third-party candidates, which is fairly common, particularly in elections that voters perceive to be somewhat close.
The other factor that stood out was the persistently large number of undecided voters in the polls, which created a level of uncertainty in what the final outcome would be on Election Day, particularly in very close states.
With these two factors, the chance of Trump winning jumped from 0.2 percent, an extreme long shot, to 77.62 percent, a favorite. Few could have foreseen this scenario to play out as it did. Election Analytics provided only one scenario for Trump to win, across a set of 21 different plausible — and in general, more likely — cases. In essence, Trump pulled a rabbit out of his hat, after everyone believed that the hat was empty.
With 20-20 hindsight, it is clear that voters overwhelmingly supported Trump, providing paths for him to take battleground states like Florida, North Carolina and Ohio, and bringing into play states like Michigan, Pennsylvania and the biggest head-scratcher of all — Wisconsin. This election provides perspective on both the strength of data and the limitations of what information one can expect to glean from data.
Until the 2018 midterms ...
Sheldon Jacobson and a team of UI students developed Election Analtics, a nonpartisan website that analyzes state polling data. For more, visit electionanalytics.cs.illinois.edu.