Why Did So Many News Outlets Fail to Predict a Trump Win? A Simple Statistics Error Helps Explain It, Says Temple University Expert

Article ID: 664565

Released: 10-Nov-2016 10:05 AM EST

Source Newsroom: Temple University

Expert Pitch

Douglas Webber, assistant professor of economics at Temple University, can help explain why many statistical models that aggregate polls got the election results so wrong.

Here is Webber's analysis:

"For simplicity's sake, say there are seven toss-up states all polling evenly 50-50 between Clinton and Trump, and that Mr. Trump needs every one of them to win the election. Intuitively, this would seem to be an overwhelmingly good situation for Sec. Clinton, with her probability of winning at least one state at about 99 percent. However, this assumes that each state's outcome is independent of every other state. But if a poll in Ohio understates Mr. Trump's support/turnout among non-college whites, then similar polls in Wisconsin, Michigan and Pennsylvania will likely do the same. Put another way, Mr. Trump did not need to get a tails on seven straight coin flips, he really only needed two."

To schedule an interview with Dr. Webber, please contact Ashwin Verghese in Temple's communications office at 215-204-7739 or ashwin@temple.edu.


Comment/Share





Chat now!